Polish reader to simplify preprocessing logic. (#112)

* Polish Reader to simplify preprocessing logic.
#  sample_transforms-> make batch -> batch_transforms in Reader.
* Clean some code
* Imporve yolov3_r50vd_dcn_obj365_pretrained_coco 41.4 to 41.8.
* Update all configs.

.pre-commit-config.yaml

@@ -11,17 +11,17 @@
     -   id: detect-private-key
         files: (?!.*paddle)^.*$
     -   id: end-of-file-fixer
-        files: \.md$
+        files: \.(md|yml)$
     -   id: trailing-whitespace
-        files: \.md$
+        files: \.(md|yml)$
 -   repo: https://github.com/Lucas-C/pre-commit-hooks
     sha: v1.0.1
     hooks:
     -   id: forbid-crlf
-        files: \.md$
+        files: \.(md|yml)$
     -   id: remove-crlf
-        files: \.md$
+        files: \.(md|yml)$
     -   id: forbid-tabs
-        files: \.md$
+        files: \.(md|yml)$
     -   id: remove-tabs
-        files: \.md$
+        files: \.(md|yml)$

configs/cascade_mask_rcnn_r50_fpn_1x.yml

 architecture: CascadeMaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 snapshot_iter: 10000
@@ -86,7 +83,7 @@ MaskAssigner:
   resolution: 28
 
 CascadeBBoxHead:
-  head: CascadeTwoFCHead 
+  head: CascadeTwoFCHead
   nms:
     keep_top_k: 100
     nms_threshold: 0.5
@@ -113,33 +110,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/cascade_rcnn_cls_aware_r101_vd_fpn_1x_softnms.yml

 architecture: CascadeRCNNClsAware
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -107,80 +104,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  sample_transforms: 
-  - !DecodeImage
-    to_rgb: True
-    with_mixup: False
-  - !NormalizeImage
-    is_channel_first: false
-    is_scale: True
-    mean:
-    - 0.485
-    - 0.456
-    - 0.406
-    std:
-    - 0.229
-    - 0.224
-    - 0.225 
-  - !ResizeImage
-    interp: 1
-    target_size:
-    - 800
-    max_size: 1333
-    use_cv2: true
-  - !Permute
-    to_bgr: false
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  sample_transforms: 
-  - !DecodeImage
-    to_rgb: True
-    with_mixup: False
-  - !NormalizeImage
-    is_channel_first: false
-    is_scale: True
-    mean:
-    - 0.485
-    - 0.456
-    - 0.406
-    std:
-    - 0.229
-    - 0.224
-    - 0.225 
-  - !ResizeImage
-    interp: 1
-    target_size:
-    - 800
-    max_size: 1333
-    use_cv2: true
-  - !Permute
-    to_bgr: false
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
\ No newline at end of file

configs/cascade_rcnn_r50_fpn_1x.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -104,34 +101,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/cascade_rcnn_r50_fpn_1x_ms_test.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -111,23 +108,22 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
 
-FasterRCNNEvalFeed:
+EvalReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+    multi_scale: true
+    num_scales: 18
+    use_flip: true
   dataset:
+    !COCODataSet
     dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    anno_path: annotations/instances_val2017.json
     image_dir: val2017
   sample_transforms:
   - !DecodeImage
@@ -160,18 +156,6 @@ FasterRCNNEvalFeed:
   - !Permute
     channel_first: true
     to_bgr: false
-  batch_transforms:
-  - !PadMSTest
-    pad_to_stride: 32
-  num_scale: 18
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
+  - !PadMultiScaleTest
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+  worker_num: 2

configs/dcn/cascade_mask_rcnn_dcnv2_se154_vd_fpn_gn_s1x.yml

 architecture: CascadeMaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 300000
 snapshot_iter: 10
 use_gpu: true
 log_iter: 20
 log_smooth_window: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/SENet154_vd_caffe_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/SENet154_vd_caffe_pretrained.tar
 weights: output/cascade_mask_rcnn_dcn_se154_vd_fpn_gn_s1x/model_final/
 metric: COCO
 num_classes: 81
@@ -96,7 +93,7 @@ MaskAssigner:
   resolution: 28
 
 CascadeBBoxHead:
-  head: CascadeXConvNormHead 
+  head: CascadeXConvNormHead
   nms:
     keep_top_k: 100
     nms_threshold: 0.5
@@ -123,17 +120,19 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
+TrainReader:
   # batch size per device
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd', 'gt_mask']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/coco
     image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  sample_transforms: 
+    anno_path: annotations/instances_train2017.json
+  sample_transforms:
   - !DecodeImage
-    to_rgb: False
-    with_mixup: False
+    to_rgb: false
   - !RandomFlipImage
     is_mask_flip: true
     is_normalized: false
@@ -142,13 +141,13 @@ MaskRCNNTrainFeed:
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !ResizeImage
     interp: 1
     target_size:
@@ -161,7 +160,7 @@ MaskRCNNTrainFeed:
     - 608
     - 640
     - 672
-    - 704 
+    - 704
     - 736
     - 768
     - 800
@@ -192,29 +191,32 @@ MaskRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  num_workers: 8
+  worker_num: 8
+  shuffle: true
 
-MaskRCNNEvalFeed:
+EvalReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    anno_path: annotations/instances_val2017.json
     image_dir: val2017
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: False
-    with_mixup: False
   - !NormalizeImage
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !ResizeImage
     interp: 1
     target_size:
@@ -227,16 +229,20 @@ MaskRCNNEvalFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  num_workers: 2
+  worker_num: 2
+  drop_empty: false
 
-MaskRCNNTestFeed:
+TestReader:
   batch_size: 1
+  inputs_def:
+    image_shape: [3,800,1333]
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
   sample_transforms:
   - !DecodeImage
     to_rgb: False
-    with_mixup: False
   - !NormalizeImage
     is_channel_first: false
     is_scale: False
@@ -254,4 +260,4 @@ MaskRCNNTestFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  num_workers: 2
+  worker_num: 2

configs/dcn/cascade_mask_rcnn_dcnv2_se154_vd_fpn_gn_s1x_ms_test.yml

 architecture: CascadeMaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 300000
 snapshot_iter: 10000
 use_gpu: true
 log_iter: 20
 log_smooth_window: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/SENet154_vd_caffe_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/SENet154_vd_caffe_pretrained.tar
 weights: output/cascade_mask_rcnn_dcn_se154_vd_fpn_gn_s1x/model_final/
 metric: COCO
 num_classes: 81
@@ -96,7 +93,7 @@ MaskAssigner:
   resolution: 28
 
 CascadeBBoxHead:
-  head: CascadeXConvNormHead 
+  head: CascadeXConvNormHead
   nms:
     keep_top_k: 100
     nms_threshold: 0.5
@@ -130,14 +127,17 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
+TrainReader:
   # batch size per device
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd', 'gt_mask']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/coco
     image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  sample_transforms: 
+    anno_path: annotations/instances_train2017.json
+  sample_transforms:
   - !DecodeImage
     to_rgb: False
     with_mixup: False
@@ -149,13 +149,13 @@ MaskRCNNTrainFeed:
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !ResizeImage
     interp: 1
     target_size:
@@ -168,7 +168,7 @@ MaskRCNNTrainFeed:
     - 608
     - 640
     - 672
-    - 704 
+    - 704
     - 736
     - 768
     - 800
@@ -199,28 +199,36 @@ MaskRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  num_workers: 8
+  worker_num: 8
+  shuffle: true
 
-MaskRCNNEvalFeed:
+EvalReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+    multi_scale: true
+    # num_scale = (len(target_size) + 1) * (1 + use_flip)
+    num_scales: 18
+    use_flip: true
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    !COCODataSet
     image_dir: val2017
-  sample_transforms: 
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
   - !DecodeImage
     to_rgb: False
   - !NormalizeImage
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !MultiscaleTestResize
     origin_target_size: 800
     origin_max_size: 1333
@@ -239,16 +247,17 @@ MaskRCNNEvalFeed:
     channel_first: true
     to_bgr: false
   batch_transforms:
-  - !PadMSTest
+  - !PadMultiScaleTest
     pad_to_stride: 32
-  # num_scale = (len(target_size) + 1) * (1 + use_flip)
-  num_scale: 18 
-  num_workers: 2
+  worker_num: 2
 
-MaskRCNNTestFeed:
+TestReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
   sample_transforms:
   - !DecodeImage
     to_rgb: False
@@ -269,4 +278,3 @@ MaskRCNNTestFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  num_workers: 2

configs/dcn/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 460000
 snapshot_iter: 10000
 use_gpu: true
@@ -109,16 +106,18 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+TrainReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
   dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
     dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
-    to_rgb: True
-    with_mixup: False
+    to_rgb: true
   - !RandomFlipImage
     prob: 0.5
   - !NormalizeImage
@@ -131,7 +130,7 @@ FasterRCNNTrainFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size: [416, 448, 480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800, 832, 864, 896, 928, 960, 992, 1024, 1056, 1088, 1120, 1152, 1184, 1216, 1248, 1280, 1312, 1344, 1376, 1408]
@@ -142,16 +141,19 @@ FasterRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+  worker_num: 2
+  shuffle: true
 
-FasterRCNNEvalFeed:
+EvalReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    !COCODataSet
     image_dir: val2017
-  sample_transforms: 
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
   - !DecodeImage
     to_rgb: True
     with_mixup: False
@@ -165,7 +167,7 @@ FasterRCNNEvalFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size:
@@ -177,13 +179,34 @@ FasterRCNNEvalFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
+  worker_num: 2
 
-FasterRCNNTestFeed:
-  batch_size: 1
+TestReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+    use_padded_im_info: true
+  batch_size: 1
+  worker_num: 2

configs/dcn/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml

 architecture: CascadeRCNNClsAware
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 460000
 snapshot_iter: 10000
 use_gpu: true
@@ -109,16 +106,17 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  batch_size: 1
+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
+    !COCODataSet
     image_dir: train2017
-  sample_transforms: 
+    anno_path: annotations/instances_train2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
   - !DecodeImage
-    to_rgb: True
-    with_mixup: False
+    to_rgb: true
   - !RandomFlipImage
     prob: 0.5
   - !NormalizeImage
@@ -131,7 +129,7 @@ FasterRCNNTrainFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size: [416, 448, 480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800, 832, 864, 896, 928, 960, 992, 1024, 1056, 1088, 1120, 1152, 1184, 1216, 1248, 1280, 1312, 1344, 1376, 1408]
@@ -142,16 +140,20 @@ FasterRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
+  batch_size: 1
+  shuffle: true
   drop_last: false
-  num_workers: 2
+  worker_num: 2
 
-FasterRCNNEvalFeed:
-  batch_size: 1
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    !COCODataSet
     image_dir: val2017
-  sample_transforms: 
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
   - !DecodeImage
     to_rgb: True
     with_mixup: False
@@ -165,7 +167,7 @@ FasterRCNNEvalFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size:
@@ -177,13 +179,37 @@ FasterRCNNEvalFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-
-FasterRCNNTestFeed:
   batch_size: 1
+  worker_num: 2
+  drop_empty: false
+
+TestReader:
+  inputs_def:
+    image_shape: [3,800,1333]
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+    use_padded_im_info: true
+  batch_size: 1
+  worker_num: 2

configs/dcn/cascade_rcnn_dcn_r101_vd_fpn_1x.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/dcn/cascade_rcnn_dcn_r50_fpn_1x.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/dcn/cascade_rcnn_dcn_x101_vd_64x4d_fpn_1x.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -26,7 +23,7 @@ ResNeXt:
   depth: 101
   feature_maps: [2, 3, 4, 5]
   freeze_at: 2
-  group_width: 4                                                                                                                                          
+  group_width: 4
   groups: 64
   variant: d
   dcn_v2_stages: [3, 4, 5]
@@ -108,34 +105,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/dcn/faster_rcnn_dcn_r101_vd_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,7 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
   # batch size per device
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2

configs/dcn/faster_rcnn_dcn_r50_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 use_gpu: true
 snapshot_iter: 10000
@@ -105,34 +102,8 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
 
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
+  # batch size per device
+  batch_size: 2

configs/dcn/faster_rcnn_dcn_r50_vd_fpn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,7 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
   # batch size per device
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2

configs/dcn/faster_rcnn_dcn_x101_vd_64x4d_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -108,36 +105,5 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-  shuffle: true
 
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-  shuffle: false
+_READER_: '../faster_fpn_reader.yml'

configs/dcn/mask_rcnn_dcn_r101_vd_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
 log_smooth_window: 20
 log_iter: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_vd_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_vd_pretrained.tar
 weights: output/mask_rcnn_dcn_r101_vd_fpn_1x/model_final
 metric: COCO
 num_classes: 81
@@ -114,33 +111,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: '../mask_fpn_reader.yml'

configs/dcn/mask_rcnn_dcn_r50_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 snapshot_iter: 10000
@@ -113,33 +110,5 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
 
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: '../mask_fpn_reader.yml'

configs/dcn/mask_rcnn_dcn_r50_vd_fpn_2x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 360000
 snapshot_iter: 10000
@@ -114,34 +111,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: '../mask_fpn_reader.yml'

configs/dcn/mask_rcnn_dcn_x101_vd_64x4d_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
 log_smooth_window: 20
 log_iter: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar
 weights: output/mask_rcnn_dcn_x101_vd_64x4d_fpn_1x/model_final
 metric: COCO
 num_classes: 81
@@ -116,33 +113,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: '../mask_fpn_reader.yml'

configs/dcn/yolov3_r50vd_dcn.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 500000
 log_smooth_window: 20
 save_dir: output
 snapshot_iter: 20000
 metric: COCO
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar
 weights: output/yolov3_r50vd_dcn/model_final
 num_classes: 80
 
@@ -62,67 +59,4 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  sample_transforms:
-    - !DecodeImage
-      to_rgb: True
-      with_mixup: True
-    - !MixupImage
-      alpha: 1.5
-      beta: 1.5
-    - !NormalizeBox {}
-    - !RandomDistort {}
-    - !ExpandImage
-      max_ratio: 4
-      prob: 0.5
-      mean:
-      - 123.675
-      - 116.28
-      - 103.53
-    - !CropImage
-      batch_sampler: [[1, 1, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.1, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.3, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.5, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.7, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.9, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.0, 1.0]]
-    - !RandomInterpImage
-      target_size: 608
-    - !RandomFlipImage
-      is_normalized: True
-    - !NormalizeImage
-      mean:
-      - 0.485
-      - 0.456
-      - 0.406
-      std:
-      - 0.229
-      - 0.224
-      - 0.225
-      is_scale: True
-      is_channel_first: False      
-    - !Permute
-      to_bgr: False
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: '../yolov3_reader.yml'

configs/dcn/yolov3_r50vd_dcn_obj365_pretrained_coco.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 55000
 log_smooth_window: 20
 save_dir: output
 snapshot_iter: 10000
 metric: COCO
-pretrain_weights: https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_obj365_pretrained.tar 
+pretrain_weights: https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_obj365_pretrained.tar
 weights: output/yolov3_r50vd_dcn_obj365_pretrained_coco/model_final
 num_classes: 80
 
@@ -62,98 +59,97 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
+TrainReader:
+  inputs_def:
+    fields: ['image', 'gt_bbox', 'gt_class', 'gt_score']
+    num_max_boxes: 50
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
+    !COCODataSet
     image_dir: train2017
+    anno_path: annotations/instances_train2017.json
+    dataset_dir: dataset/coco
+    with_background: false
   sample_transforms:
     - !DecodeImage
       to_rgb: True
-      with_mixup: False
-    - !NormalizeBox {}
-    - !CropImage
-      batch_sampler: [[1, 1, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.1, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.3, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.5, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.7, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.9, 1.0],
-        [1, 50, 0.3, 1.0, 0.5, 2.0, 0.0, 1.0]]
-    - !RandomInterpImage
-      target_size: 608
+    - !RandomCrop {}
     - !RandomFlipImage
-      is_normalized: True
+      is_normalized: false
+    - !NormalizeBox {}
+    - !PadBox
+      num_max_boxes: 50
+    - !BboxXYXY2XYWH {}
+  batch_transforms:
+    - !RandomShape
+      sizes: [320, 352, 384, 416, 448, 480, 512, 544, 576, 608]
+      random_inter: True
     - !NormalizeImage
-      mean:
-      - 0.485
-      - 0.456
-      - 0.406
-      std:
-      - 0.229
-      - 0.224
-      - 0.225
-      is_scale: False
-      is_channel_first: False      
+      mean: [0.485, 0.456, 0.406]
+      std: [0.229, 0.224, 0.225]
+      is_scale: True
+      is_channel_first: false
     - !Permute
-      to_bgr: False
-  num_workers: 8
-  bufsize: 128
+      to_bgr: false
+      channel_first: True
+  batch_size: 8
+  shuffle: true
+  drop_last: true
+  worker_num: 8
+  bufsize: 32
   use_process: true
 
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
+EvalReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'im_size', 'im_id']
+    num_max_boxes: 50
   dataset:
+    !COCODataSet
     dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    anno_path: annotations/instances_val2017.json
     image_dir: val2017
+    with_background: false
   sample_transforms:
     - !DecodeImage
       to_rgb: True
-      with_mixup: False
+      with_mixup: false
     - !ResizeImage
-      interp: 2 
+      interp: 2
       target_size: 608
     - !NormalizeImage
-      mean:
-      - 0.485
-      - 0.456
-      - 0.406
-      std:
-      - 0.229
-      - 0.224
-      - 0.225
-      is_scale: False
-      is_channel_first: False      
+      mean: [0.485, 0.456, 0.406]
+      std: [0.229, 0.224, 0.225]
+      is_scale: True
+      is_channel_first: false
     - !Permute
-      to_bgr: False
-
+      to_bgr: false
+      channel_first: True
+  batch_size: 8
+  drop_empty: false
+  worker_num: 8
+  bufsize: 32
 
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
+TestReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'im_size', 'im_id']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+    !ImageFolder
+      anno_path: annotations/instances_val2017.json
+      with_background: false
   sample_transforms:
     - !DecodeImage
       to_rgb: True
-      with_mixup: False
+      with_mixup: false
     - !ResizeImage
-      interp: 2 
+      interp: 2
       target_size: 608
     - !NormalizeImage
-      mean:
-      - 0.485
-      - 0.456
-      - 0.406
-      std:
-      - 0.229
-      - 0.224
-      - 0.225
-      is_scale: False
-      is_channel_first: False      
+      mean: [0.485, 0.456, 0.406]
+      std: [0.229, 0.224, 0.225]
+      is_scale: True
+      is_channel_first: false
     - !Permute
-      to_bgr: False
-
+      to_bgr: false
+      channel_first: True
+  batch_size: 1

configs/face_detection/blazeface.yml

 architecture: BlazeFace
 max_iters: 320000
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
-pretrain_weights: 
+pretrain_weights:
 use_gpu: true
 snapshot_iter: 10000
 log_smooth_window: 20
@@ -43,18 +40,18 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 8
-  use_process: True
+TrainReader:
+  inputs_def:
+    image_shape: [3, 640, 640]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_train_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_train_bbx_gt.txt
     image_dir: WIDER_train/images
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
@@ -82,49 +79,41 @@ SSDTrainFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
+  batch_size: 8
+  use_process: true
+  shuffle: true
 
-SSDEvalFeed:
-  batch_size: 1
-  use_process: false
-  fields: ['image', 'im_id', 'gt_box']
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_id']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_val_bbx_gt.txt   
+    anno_path: wider_face_split/wider_face_val_bbx_gt.txt
     image_dir: WIDER_val/images
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
-  - !ResizeImage
-    interp: 1
-    target_size: 640
-    use_cv2: false
   - !Permute {}
   - !NormalizeImage
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
-
-SSDTestFeed:
   batch_size: 1
-  use_process: false
+
+TestReader:
+  inputs_def:
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !ResizeImage
-    interp: 1
-    target_size: 640
-    use_cv2: false
   - !Permute {}
   - !NormalizeImage
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
+  batch_size: 1

configs/face_detection/blazeface_nas.yml

 architecture: BlazeFace
 max_iters: 320000
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
-pretrain_weights: 
+pretrain_weights:
 use_gpu: true
 snapshot_iter: 10000
 log_smooth_window: 20
@@ -45,18 +42,18 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 8
-  use_process: True
+TrainReader:
+  inputs_def:
+    image_shape: [3, 640, 640]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_train_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_train_bbx_gt.txt
     image_dir: WIDER_train/images
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
@@ -84,21 +81,21 @@ SSDTrainFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
+  batch_size: 8
+  use_process: true
+  shuffle: true
 
-SSDEvalFeed:
-  batch_size: 1
-  use_process: false
-  fields: ['image', 'im_id', 'gt_box']
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_id']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_val_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_val_bbx_gt.txt
     image_dir: WIDER_val/images
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
   - !ResizeImage
     interp: 1
@@ -109,18 +106,18 @@ SSDEvalFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
-
-SSDTestFeed:
   batch_size: 1
-  use_process: false
+
+TestReader:
+  inputs_def:
+    image_shape: [3,640,640]
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !ResizeImage
     interp: 1
     target_size: 640
@@ -130,3 +127,4 @@ SSDTestFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
+  batch_size: 1

configs/face_detection/faceboxes.yml

 architecture: FaceBoxes
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 pretrain_weights:
 use_gpu: true
 max_iters: 320000
@@ -43,18 +40,21 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
+TrainReader:
   batch_size: 8
   use_process: True
+  shuffle: true
+  inputs_def:
+    image_shape: [3, 640, 640]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_train_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_train_bbx_gt.txt
     image_dir: WIDER_train/images
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
@@ -83,48 +83,37 @@ SSDTrainFeed:
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
 
-SSDEvalFeed:
+EvalReader:
   batch_size: 1
   use_process: false
-  fields: ['image', 'im_id', 'gt_box']
+  inputs_def:
+    fields: ['image', 'im_id']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_val_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_val_bbx_gt.txt
     image_dir: WIDER_val/images
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !NormalizeBox {}
-  - !ResizeImage
-    interp: 1
-    target_size: 640
-    use_cv2: false
   - !Permute {}
   - !NormalizeImage
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
 
-SSDTestFeed:
-  batch_size: 1
-  use_process: false
+TestReader:
+  inputs_def:
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !ResizeImage
-    interp: 1
-    target_size: 640
-    use_cv2: false
   - !Permute {}
   - !NormalizeImage
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
+  batch_size: 1

configs/face_detection/faceboxes_lite.yml

 architecture: FaceBoxes
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 pretrain_weights:
 use_gpu: true
 max_iters: 320000
@@ -43,18 +40,21 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
+TrainReader:
   batch_size: 8
   use_process: True
+  shuffle: true
+  inputs_def:
+    image_shape: [3, 640, 640]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_train_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_train_bbx_gt.txt
     image_dir: WIDER_train/images
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
@@ -83,48 +83,38 @@ SSDTrainFeed:
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
 
-SSDEvalFeed:
+EvalReader:
   batch_size: 1
   use_process: false
-  fields: ['image', 'im_id', 'gt_box']
+  inputs_def:
+    fields: ['image', 'im_id']
   dataset:
+    !WIDERFaceDataSet
     dataset_dir: dataset/wider_face
-    annotation: wider_face_split/wider_face_val_bbx_gt.txt
+    anno_path: wider_face_split/wider_face_val_bbx_gt.txt
     image_dir: WIDER_val/images
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
   - !NormalizeBox {}
-  - !ResizeImage
-    interp: 1
-    target_size: 640
-    use_cv2: false
   - !Permute {}
   - !NormalizeImage
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
 
-SSDTestFeed:
-  batch_size: 1
-  use_process: false
+TestReader:
+  inputs_def:
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 640, 640]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !ResizeImage
-    interp: 1
-    target_size: 640
-    use_cv2: false
   - !Permute {}
   - !NormalizeImage
     is_scale: false
     mean: [104, 117, 123]
     std: [127.502231, 127.502231, 127.502231]
+  batch_size: 1

configs/faster_fpn_reader.yml

+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
+  dataset:
+    !COCODataSet
+    image_dir: train2017
+    anno_path: annotations/instances_train2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !RandomFlipImage
+    prob: 0.5
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    target_size: 800
+    max_size: 1333
+    interp: 1
+    use_cv2: true
+  - !Permute
+    to_bgr: false
+    channel_first: true
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 32
+    use_padded_im_info: false
+  batch_size: 1
+  shuffle: true
+  worker_num: 2
+  use_process: false
+
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+    # for voc
+    #fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+  dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 32
+    use_padded_im_info: true
+  batch_size: 1
+  shuffle: false
+  drop_empty: false
+  worker_num: 2
+
+TestReader:
+  inputs_def:
+    # set image_shape if needed
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+  dataset:
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 32
+    use_padded_im_info: true
+  batch_size: 1
+  shuffle: false

configs/faster_rcnn_cbr101_vd_dual_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2

configs/faster_rcnn_cbr50_vd_dual_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,7 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   # batch size per device
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2

configs/faster_rcnn_r101_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 log_smooth_window: 20
@@ -91,25 +88,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'faster_reader.yml'

configs/faster_rcnn_r101_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -103,34 +100,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'faster_fpn_reader.yml'

configs/faster_rcnn_r101_fpn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 360000
 snapshot_iter: 10000
 use_gpu: true
@@ -103,34 +100,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'faster_fpn_reader.yml'

configs/faster_rcnn_r101_vd_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -104,34 +101,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'faster_fpn_reader.yml'

configs/faster_rcnn_r101_vd_fpn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 360000
 snapshot_iter: 10000
 use_gpu: true
@@ -104,34 +101,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'faster_fpn_reader.yml'

configs/faster_rcnn_r50_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 log_smooth_window: 20
@@ -91,25 +88,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'faster_reader.yml'

configs/faster_rcnn_r50_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 use_gpu: true
 max_iters: 360000
 log_smooth_window: 20
@@ -91,25 +88,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'faster_reader.yml'

configs/faster_rcnn_r50_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 use_gpu: true
 snapshot_iter: 10000
@@ -104,34 +101,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/faster_rcnn_r50_fpn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 use_gpu: true
 snapshot_iter: 10000
@@ -104,34 +101,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/faster_rcnn_r50_vd_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 log_smooth_window: 20
@@ -93,25 +90,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  drop_last: false
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'faster_reader.yml'

configs/faster_rcnn_r50_vd_fpn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -104,34 +101,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2

configs/faster_rcnn_se154_vd_fpn_s1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 260000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,34 +103,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'faster_fpn_reader.yml'

configs/faster_rcnn_x101_vd_64x4d_fpn_1x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -107,36 +104,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-  shuffle: true
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-  shuffle: false
+_READER_: 'faster_fpn_reader.yml'

configs/faster_rcnn_x101_vd_64x4d_fpn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 360000
 snapshot_iter: 10000
 use_gpu: true
@@ -106,36 +103,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-  shuffle: true
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-  shuffle: false
+_READER_: 'faster_fpn_reader.yml'

configs/faster_reader.yml

+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
+  dataset:
+    !COCODataSet
+    image_dir: train2017
+    anno_path: annotations/instances_train2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !RandomFlipImage
+    prob: 0.5
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    target_size: 800
+    max_size: 1333
+    interp: 1
+    use_cv2: true
+  - !Permute
+    to_bgr: false
+    channel_first: true
+  batch_size: 1
+  shuffle: true
+  worker_num: 2
+  use_process: false
+
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+    # for voc
+    #fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+  dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_size: 1
+  shuffle: false
+  drop_empty: false
+  worker_num: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3,800,1333]
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+  dataset:
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_size: 1
+  shuffle: false

configs/gn/cascade_mask_rcnn_r50_fpn_gn_2x.yml

 architecture: CascadeMaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -88,7 +85,7 @@ MaskAssigner:
   resolution: 28
 
 CascadeBBoxHead:
-  head: CascadeXConvNormHead 
+  head: CascadeXConvNormHead
   nms:
     keep_top_k: 100
     nms_threshold: 0.5
@@ -115,33 +112,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
+_READER_: '../mask_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2

configs/gn/faster_rcnn_r50_fpn_gn_2x.yml

 architecture: FasterRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
@@ -104,34 +101,6 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: '../faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 16
-
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  drop_last: false
-  num_workers: 2

configs/gn/mask_rcnn_r50_fpn_gn_2x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 360000
 snapshot_iter: 10000
 use_gpu: true
@@ -113,33 +110,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: '../mask_fpn_reader.yml'

configs/mask_fpn_reader.yml

+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd', 'gt_mask']
+  dataset:
+    !COCODataSet
+    image_dir: train2017
+    anno_path: annotations/instances_train2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !RandomFlipImage
+    prob: 0.5
+    is_mask_flip: true
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    target_size: 800
+    max_size: 1333
+    interp: 1
+    use_cv2: true
+  - !Permute
+    to_bgr: false
+    channel_first: true
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 32
+    use_padded_im_info: false
+  batch_size: 1
+  shuffle: true
+  worker_num: 2
+  drop_last: false
+  use_process: false
+
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+    # for voc
+    #fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+  dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 32
+    use_padded_im_info: true
+  batch_size: 1
+  shuffle: false
+  drop_last: false
+  drop_empty: false
+  worker_num: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3,800,1333]
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+  dataset:
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 32
+    use_padded_im_info: true
+  batch_size: 1
+  shuffle: false
+  drop_last: false

configs/mask_rcnn_r101_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 snapshot_iter: 10000
@@ -111,33 +108,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_r101_vd_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
 log_smooth_window: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_vd_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_vd_pretrained.tar
 weights: output/mask_rcnn_r101_vd_fpn_1x/model_final
 metric: COCO
 num_classes: 81
@@ -112,33 +109,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco 
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco 
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_r50_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 snapshot_iter: 10000
@@ -102,22 +99,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'mask_reader.yml'

configs/mask_rcnn_r50_2x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 360000
 snapshot_iter: 10000
@@ -104,22 +101,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'mask_reader.yml'

configs/mask_rcnn_r50_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 180000
 snapshot_iter: 10000
@@ -111,33 +108,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_r50_fpn_2x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 360000
 snapshot_iter: 10000
 use_gpu: true
@@ -111,33 +108,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_r50_vd_fpn_2x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 use_gpu: true
 max_iters: 360000
 snapshot_iter: 10000
@@ -112,34 +109,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_se154_vd_fpn_s1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 260000
 snapshot_iter: 10000
 use_gpu: true
@@ -114,34 +111,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  # batch size per device
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    image_dir: train2017
-    annotation: annotations/instances_train2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_x101_vd_64x4d_fpn_1x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 180000
 snapshot_iter: 10000
 use_gpu: true
 log_smooth_window: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar
 weights: output/mask_rcnn_x101_vd_64x4d_fpn_1x/model_final
 metric: COCO
 num_classes: 81
@@ -114,33 +111,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco 
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_rcnn_x101_vd_64x4d_fpn_2x.yml

 architecture: MaskRCNN
-train_feed: MaskRCNNTrainFeed
-eval_feed: MaskRCNNEvalFeed
-test_feed: MaskRCNNTestFeed
 max_iters: 360000
 snapshot_iter: 10000
 use_gpu: true
 log_smooth_window: 20
 save_dir: output
-pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar 
+pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar
 weights: output/mask_rcnn_x101_vd_64x4d_fpn_2x/model_final
 metric: COCO
 num_classes: 81
@@ -114,33 +111,4 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-MaskRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco 
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
-
-MaskRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: annotations/instances_val2017.json
-  batch_transforms:
-  - !PadBatch
-    pad_to_stride: 32
-  num_workers: 2
+_READER_: 'mask_fpn_reader.yml'

configs/mask_reader.yml

+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd', 'gt_mask']
+  dataset:
+    !COCODataSet
+    image_dir: train2017
+    anno_path: annotations/instances_train2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !RandomFlipImage
+    prob: 0.5
+    is_mask_flip: true
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    target_size: 800
+    max_size: 1333
+    interp: 1
+    use_cv2: true
+  - !Permute
+    to_bgr: false
+    channel_first: true
+  batch_size: 1
+  shuffle: true
+  worker_num: 2
+  drop_last: false
+  use_process: false
+
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+    # for voc
+    #fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+  dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_size: 1
+  shuffle: false
+  drop_last: false
+  drop_empty: false
+  worker_num: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3,800,1333]
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
+  dataset:
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
+  batch_size: 1
+  shuffle: false
+  drop_last: false

configs/obj365/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml

 architecture: CascadeRCNNClsAware
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 800000
 snapshot_iter: 10000
 use_gpu: true
@@ -110,16 +107,17 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  batch_size: 1
+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/obj365
-    annotation: train.json
+    anno_path: train.json
     image_dir: train
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: True
-    with_mixup: False
   - !RandomFlipImage
     prob: 0.5
   - !NormalizeImage
@@ -132,7 +130,7 @@ FasterRCNNTrainFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size: [416, 448, 480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800, 832, 864, 896, 928, 960, 992, 1024, 1056, 1088, 1120, 1152, 1184, 1216, 1248, 1280, 1312, 1344, 1376, 1408]
@@ -143,16 +141,20 @@ FasterRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
+  batch_size: 1
+  shuffle: true
   drop_last: false
-  num_workers: 2
+  worker_num: 2
 
-FasterRCNNEvalFeed:
-  batch_size: 1
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/obj365
-    annotation: val.json
+    anno_path: val.json
     image_dir: val
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: True
     with_mixup: False
@@ -166,7 +168,7 @@ FasterRCNNEvalFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size:
@@ -178,13 +180,36 @@ FasterRCNNEvalFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-
-FasterRCNNTestFeed:
   batch_size: 1
+  worker_num: 2
+  drop_empty: false
+
+TestReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/obj365/val.json
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+    with_mixup: false
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+    use_padded_im_info: true
+  batch_size: 1
+  worker_num: 2

configs/obj365/cascade_rcnn_dcnv2_se154_vd_fpn_gn_cas.yml

 architecture: CascadeRCNN
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 500000
 snapshot_iter: 10000
 use_gpu: true
@@ -85,7 +82,7 @@ CascadeBBoxAssigner:
   fg_fraction: 0.25
 
 CascadeBBoxHead:
-  head: CascadeXConvNormHead 
+  head: CascadeXConvNormHead
   nms:
     keep_top_k: 100
     nms_threshold: 0.5
@@ -115,16 +112,17 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  batch_size: 1
+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/objects365
-    annotation: annotations/train.json
+    anno_path: annotations/train.json
     image_dir: train
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: False
-    with_mixup: False
   - !RandomFlipImage
     is_mask_flip: true
     is_normalized: false
@@ -133,13 +131,13 @@ FasterRCNNTrainFeed:
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !ResizeImage
     interp: 1
     target_size:
@@ -152,7 +150,7 @@ FasterRCNNTrainFeed:
     - 608
     - 640
     - 672
-    - 704 
+    - 704
     - 736
     - 768
     - 800
@@ -183,30 +181,34 @@ FasterRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  num_workers: 4
+  batch_size: 1
+  worker_num: 4
+  shuffle: true
   class_aware_sampling: true
+  use_process: false
 
-FasterRCNNEvalFeed:
-  batch_size: 1
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/objects365
-    annotation: annotations/val.json
+    anno_path: annotations/val.json
     image_dir: val
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: False
-    with_mixup: False
   - !NormalizeImage
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !ResizeImage
     target_size: 800
     max_size: 1333
@@ -217,31 +219,33 @@ FasterRCNNEvalFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
+  batch_size: 1
+  drop_empty: false
+  worker_num: 2
 
-FasterRCNNTestFeed:
+TestReader:
   batch_size: 1
   dataset:
-    annotation: dataset/obj365/annotations/val.json
-  sample_transforms: 
+    !ImageFolder
+    anno_path: dataset/obj365/annotations/val.json
+  sample_transforms:
   - !DecodeImage
     to_rgb: False
-    with_mixup: False
   - !NormalizeImage
     is_channel_first: false
     is_scale: False
     mean:
-    - 102.9801 
+    - 102.9801
     - 115.9465
     - 122.7717
     std:
-    - 1.0 
-    - 1.0 
-    - 1.0 
+    - 1.0
+    - 1.0
+    - 1.0
   - !Permute
     channel_first: true
     to_bgr: false
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+  worker_num: 2

configs/oidv5/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml

 architecture: CascadeRCNNClsAware
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 1500000
 snapshot_iter: 10000
 use_gpu: true
@@ -109,16 +106,17 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  batch_size: 1
+TrainReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/oid
-    annotation: train.json
+    anno_path: train.json
     image_dir: train
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: True
-    with_mixup: False
   - !RandomFlipImage
     prob: 0.5
   - !NormalizeImage
@@ -131,7 +129,7 @@ FasterRCNNTrainFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size: [416, 448, 480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800, 832, 864, 896, 928, 960, 992, 1024, 1056, 1088, 1120, 1152, 1184, 1216, 1248, 1280, 1312, 1344, 1376, 1408]
@@ -142,16 +140,20 @@ FasterRCNNTrainFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
+  batch_size: 1
   drop_last: false
-  num_workers: 2
+  shuffle: true
+  worker_num: 2
 
-FasterRCNNEvalFeed:
-  batch_size: 1
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
+    !COCODataSet
     dataset_dir: dataset/oidv5
-    annotation: val.json
+    anno_path: val.json
     image_dir: val
-  sample_transforms: 
+  sample_transforms:
   - !DecodeImage
     to_rgb: True
     with_mixup: False
@@ -165,7 +167,7 @@ FasterRCNNEvalFeed:
     std:
     - 0.229
     - 0.224
-    - 0.225 
+    - 0.225
   - !ResizeImage
     interp: 1
     target_size:
@@ -177,13 +179,34 @@ FasterRCNNEvalFeed:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
+  batch_size: 1
+  worker_num: 2
+  drop_empty: false
 
-FasterRCNNTestFeed:
+TestReader:
   batch_size: 1
+  inputs_def:
+    fields: ['image', 'im_info', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/oidv5/val.json
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !NormalizeImage
+    is_channel_first: false
+    is_scale: true
+    mean: [0.485,0.456,0.406]
+    std: [0.229, 0.224,0.225]
+  - !ResizeImage
+    interp: 1
+    max_size: 1333
+    target_size: 800
+    use_cv2: true
+  - !Permute
+    channel_first: true
+    to_bgr: false
   batch_transforms:
   - !PadBatch
     pad_to_stride: 32
-  drop_last: false
-  num_workers: 2
+  worker_num: 2

configs/retinanet_r101_fpn_1x.yml

 architecture: RetinaNet
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 use_gpu: true
 pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_pretrained.tar
@@ -73,33 +70,21 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  num_workers: 2
 
-FasterRCNNEvalFeed:
+EvalReader:
   batch_size: 2
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-  num_workers: 2
 
-FasterRCNNTestFeed:
+TestReader:
   batch_size: 1
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  num_workers: 2

configs/retinanet_r50_fpn_1x.yml

 architecture: RetinaNet
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 90000
 use_gpu: true
 pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar
@@ -73,33 +70,21 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  num_workers: 2
 
-FasterRCNNEvalFeed:
+EvalReader:
   batch_size: 2
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  num_workers: 2
 
-FasterRCNNTestFeed:
+TestReader:
   batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  num_workers: 2

configs/retinanet_x101_vd_64x4d_fpn_1x.yml

 architecture: RetinaNet
-train_feed: FasterRCNNTrainFeed
-eval_feed: FasterRCNNEvalFeed
-test_feed: FasterRCNNTestFeed
 max_iters: 180000
 use_gpu: true
 pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNeXt101_vd_64x4d_pretrained.tar
@@ -76,33 +73,18 @@ OptimizerBuilder:
     factor: 0.0001
     type: L2
 
-FasterRCNNTrainFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
+_READER_: 'faster_fpn_reader.yml'
+TrainReader:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  num_workers: 2
 
-FasterRCNNEvalFeed:
-  batch_size: 1
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
+EvalReader:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  num_workers: 2
 
-FasterRCNNTestFeed:
-  batch_size: 1
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+TestReader:
   batch_transforms:
   - !PadBatch
     pad_to_stride: 128
-  num_workers: 2

configs/ssd/ssd_mobilenet_v1_voc.yml

 architecture: SSD
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 pretrain_weights: https://paddlemodels.bj.bcebos.com/object_detection/ssd_mobilenet_v1_coco_pretrained.tar
 use_gpu: true
 max_iters: 28000
@@ -56,25 +53,91 @@ OptimizerBuilder:
     factor: 0.00005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 32
-  use_process: true
+TrainReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !VOCDataSet
+    anno_path: trainval.txt
     dataset_dir: dataset/voc
-    annotation: trainval.txt
     use_default_label: true
-
-SSDEvalFeed:
-  batch_size: 64
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !RandomDistort
+    brightness_lower: 0.875
+    brightness_upper: 1.125
+    is_order: true
+  - !RandomExpand
+    fill_value: [127.5, 127.5, 127.5]
+  - !RandomCrop
+    allow_no_crop: false
+  - !NormalizeBox {}
+  - !ResizeImage
+    interp: 1
+    target_size: 300
+    use_cv2: false
+  - !RandomFlipImage
+    is_normalized: true
+  - !Permute {}
+  - !NormalizeImage
+    is_scale: false
+    mean: [127.5, 127.5, 127.5]
+    std: [127.502231, 127.502231, 127.502231]
+  batch_size: 32
+  shuffle: true
+  drop_last: true
+  worker_num: 8
+  bufsize: 16
   use_process: true
+
+EvalReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'gt_bbox', 'gt_class', 'im_shape', 'im_id', 'is_difficult']
   dataset:
+    !VOCDataSet
+    anno_path: test.txt
     dataset_dir: dataset/voc
-    annotation: test.txt
     use_default_label: true
-  drop_last: false
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !NormalizeBox {}
+  - !ResizeImage
+    interp: 1
+    target_size: 300
+    use_cv2: false
+  - !Permute {}
+  - !NormalizeImage
+    is_scale: false
+    mean: [127.5, 127.5, 127.5]
+    std: [127.502231, 127.502231, 127.502231]
+  batch_size: 32
+  worker_num: 8
+  bufsize: 32
+  use_process: false
 
-SSDTestFeed:
-  batch_size: 1
+TestReader:
+  inputs_def:
+    image_shape: [3,300,300]
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
+    anno_path: test.txt
     use_default_label: true
-  drop_last: false
+  sample_transforms:
+  - !DecodeImage
+    to_rgb: true
+  - !ResizeImage
+    interp: 1
+    max_size: 0
+    target_size: 300
+    use_cv2: false
+  - !Permute {}
+  - !NormalizeImage
+    is_scale: false
+    mean: [127.5, 127.5, 127.5]
+    std: [127.502231, 127.502231, 127.502231]
+  batch_size: 1

configs/ssd/ssd_vgg16_300.yml

 architecture: SSD
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 use_gpu: true
 max_iters: 400000
 snapshot_iter: 10000
@@ -60,37 +57,27 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 8
+TrainReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
     dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  image_shape: [3, 300, 300]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
     brightness_upper: 1.125
     is_order: true
-  - !ExpandImage
-    max_ratio: 4
-    mean: [104, 117, 123]
-    prob: 0.5
-  - !CropImage
-    avoid_no_bbox: true
-    batch_sampler:
-    - [1, 1, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.1, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.3, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.5, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.7, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.9, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.0, 1.0]
-    satisfy_all: false
+  - !RandomExpand
+    fill_value: [104, 117, 123]
+  - !RandomCrop
+    allow_no_crop: true
+  - !NormalizeBox {}
   - !ResizeImage
     interp: 1
     target_size: 300
@@ -103,19 +90,27 @@ SSDTrainFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 8
+  shuffle: true
+  worker_num: 8
+  bufsize: 32
+  use_process: true
+  drop_empty: true
 
-SSDEvalFeed:
-  batch_size: 16
+EvalReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'gt_bbox', 'gt_class', 'im_shape', 'im_id', 'is_difficult']
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    !COCODataSet
     image_dir: val2017
-  drop_last: false
-  image_shape: [3, 300, 300]
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
   sample_transforms:
   - !DecodeImage
     to_rgb: true
     with_mixup: false
+  - !NormalizeBox {}
   - !ResizeImage
     interp: 1
     target_size: 300
@@ -126,12 +121,17 @@ SSDEvalFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 16
+  worker_num: 8
+  bufsize: 32
 
-SSDTestFeed:
-  batch_size: 1
+TestReader:
+  inputs_def:
+    image_shape: [3,300,300]
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  image_shape: [3, 300, 300]
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -147,3 +147,4 @@ SSDTestFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 1

configs/ssd/ssd_vgg16_300_voc.yml

 architecture: SSD
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 use_gpu: true
 max_iters: 120001
 snapshot_iter: 10000
@@ -60,41 +57,31 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 8
+TrainReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: trainval.txt
+    anno_path: trainval.txt
     use_default_label: true
-  image_shape: [3, 300, 300]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
     brightness_upper: 1.125
     is_order: true
-  - !ExpandImage
-    max_ratio: 4
-    mean: [104, 117, 123]
-    prob: 0.5
-  - !CropImage
-    avoid_no_bbox: true
-    batch_sampler:
-    - [1, 1, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.1, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.3, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.5, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.7, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.9, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.0, 1.0]
-    satisfy_all: false
+  - !RandomExpand
+    fill_value: [104, 117, 123]
+  - !RandomCrop
+    allow_no_crop: true
+  - !NormalizeBox {}
   - !ResizeImage
     interp: 1
     target_size: 300
-    use_cv2: False
+    use_cv2: false
   - !RandomFlipImage
     is_normalized: true
   - !Permute
@@ -103,15 +90,21 @@ SSDTrainFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 8
+  shuffle: true
+  worker_num: 8
+  bufsize: 32
+  use_process: 8
 
-SSDEvalFeed:
-  batch_size: 32
+EvalReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'gt_bbox', 'gt_class', 'im_shape', 'im_id', 'is_difficult']
   dataset:
+    !VOCDataSet
+    anno_path: test.txt
     dataset_dir: dataset/voc
-    annotation: test.txt
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 300, 300]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -125,15 +118,20 @@ SSDEvalFeed:
     to_bgr: false
   - !NormalizeImage
     is_scale: false
-    mean: [104, 117, 123]  
+    mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 32
+  worker_num: 8
+  bufsize: 32
 
-SSDTestFeed:
-  batch_size: 1
+TestReader:
+  inputs_def:
+    image_shape: [3,300,300]
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
+    anno_path: test.txt
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 300, 300]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -149,3 +147,4 @@ SSDTestFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 1

configs/ssd/ssd_vgg16_512.yml

 architecture: SSD
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 use_gpu: true
 max_iters: 400000
 snapshot_iter: 10000
@@ -62,37 +59,28 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 8
+TrainReader:
+  inputs_def:
+    image_shape: [3, 512, 512]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !COCODataSet
+    image_dir: val2017
+    anno_path: annotations/instances_val2017.json
     dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  image_shape: [3, 512, 512]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
     with_mixup: false
-  - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
     brightness_upper: 1.125
     is_order: true
-  - !ExpandImage
-    max_ratio: 4
-    mean: [104, 117, 123]
-    prob: 0.5
-  - !CropImage
-    avoid_no_bbox: true
-    batch_sampler:
-    - [1, 1, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.1, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.3, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.5, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.7, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.9, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.0, 1.0]
-    satisfy_all: false
+  - !RandomExpand
+    fill_value: [104, 117, 123]
+  - !RandomCrop
+    allow_no_crop: true
+  - !NormalizeBox {}
   - !ResizeImage
     interp: 1
     target_size: 512
@@ -105,15 +93,21 @@ SSDTrainFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
-
-SSDEvalFeed:
   batch_size: 8
+  shuffle: true
+  worker_num: 8
+  bufsize: 32
+  use_process: 8
+
+EvalReader:
+  inputs_def:
+    image_shape: [3,512,512]
+    fields: ['image', 'gt_bbox', 'gt_class', 'im_shape', 'im_id', 'is_difficult']
   dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
+    !COCODataSet
     image_dir: val2017
-  drop_last: false
-  image_shape: [3, 512, 512]
+    anno_path: annotations/instances_val2017.json
+    dataset_dir: dataset/coco
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -128,12 +122,18 @@ SSDEvalFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 8
+  worker_num: 8
+  bufsize: 32
+  drop_empty: false
 
-SSDTestFeed:
-  batch_size: 1
+TestReader:
+  inputs_def:
+    image_shape: [3,512,512]
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
-  image_shape: [3, 512, 512]
+    !ImageFolder
+    anno_path: annotations/instances_val2017.json
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -149,3 +149,4 @@ SSDTestFeed:
     is_scale: false
     mean: [104, 117, 123]
     std: [1, 1, 1]
+  batch_size: 1

configs/ssd/ssd_vgg16_512_voc.yml

 architecture: SSD
-train_feed: SSDTrainFeed
-eval_feed: SSDEvalFeed
-test_feed: SSDTestFeed
 use_gpu: true
 max_iters: 120000
 snapshot_iter: 10000
@@ -64,37 +61,27 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-SSDTrainFeed:
-  batch_size: 8
+TrainReader:
+  inputs_def:
+    image_shape: [3, 512, 512]
+    fields: ['image', 'gt_bbox', 'gt_class']
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: trainval.txt
+    anno_path: trainval.txt
     use_default_label: true
-  image_shape: [3, 512, 512]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
-    with_mixup: false
-  - !NormalizeBox {}
   - !RandomDistort
     brightness_lower: 0.875
     brightness_upper: 1.125
     is_order: true
-  - !ExpandImage
-    max_ratio: 4
-    mean: [123, 117, 104]
-    prob: 0.5
-  - !CropImage
-    avoid_no_bbox: true
-    batch_sampler:
-    - [1, 1, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.1, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.3, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.5, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.7, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.9, 0.0]
-    - [1, 50, 0.3, 1.0, 0.5, 2.0, 0.0, 1.0]
-    satisfy_all: false
+  - !RandomExpand
+    fill_value: [123, 117, 104]
+  - !RandomCrop
+    allow_no_crop: true
+  - !NormalizeBox {}
   - !ResizeImage
     interp: 1
     target_size: 512
@@ -107,15 +94,21 @@ SSDTrainFeed:
     is_scale: false
     mean: [123, 117, 104]
     std: [1, 1, 1]
+  batch_size: 8
+  shuffle: true
+  worker_num: 8
+  bufsize: 32
+  use_process: 8
 
-SSDEvalFeed:
-  batch_size: 32
+EvalReader:
+  inputs_def:
+    image_shape: [3, 512, 512]
+    fields: ['image', 'gt_bbox', 'gt_class', 'im_shape', 'im_id', 'is_difficult']
   dataset:
+    !VOCDataSet
+    anno_path: test.txt
     dataset_dir: dataset/voc
-    annotation: test.txt
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 512, 512]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -131,13 +124,18 @@ SSDEvalFeed:
     is_scale: false
     mean: [123, 117, 104]
     std: [1, 1, 1]
+  batch_size: 32
+  worker_num: 8
+  bufsize: 32
 
-SSDTestFeed:
-  batch_size: 1
+TestReader:
+  inputs_def:
+    image_shape: [3,512,512]
+    fields: ['image', 'im_id', 'im_shape']
   dataset:
+    !ImageFolder
+    anno_path: test.txt
     use_default_label: true
-  drop_last: false
-  image_shape: [3, 512, 512]
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -153,3 +151,4 @@ SSDTestFeed:
     is_scale: false
     mean: [123, 117, 104]
     std: [1, 1, 1]
+  batch_size: 1

configs/yolov3_darknet.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 500200
 log_smooth_window: 20
@@ -57,26 +54,4 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'yolov3_reader.yml'

configs/yolov3_darknet_voc.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 70000
 log_smooth_window: 20
@@ -58,27 +55,31 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
+_READER_: 'yolov3_reader.yml'
+TrainReader:
+  inputs_def:
+    fields: ['image', 'gt_bbox', 'gt_class', 'gt_score']
+    num_max_boxes: 50
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: trainval.txt
+    anno_path: trainval.txt
     use_default_label: true
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-  mixup_epoch: 250
+    with_background: false
 
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_size', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+    num_max_boxes: 50
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: test.txt
+    anno_path: test.txt
     use_default_label: true
+    with_background: false
 
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
+TestReader:
   dataset:
+    !ImageFolder
     use_default_label: true
+    with_background: false

configs/yolov3_mobilenet_v1.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 500200
 log_smooth_window: 20
@@ -58,26 +55,4 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'yolov3_reader.yml'

configs/yolov3_mobilenet_v1_fruit.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 20000
 log_smooth_window: 20
 save_dir: output
-snapshot_iter: 1000
+snapshot_iter: 200
 metric: VOC
 map_type: 11point
 pretrain_weights: https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar
@@ -60,16 +57,19 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 1
+_READER_: 'yolov3_reader.yml'
+# will merge TrainReader into yolov3_reader.yml
+TrainReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'gt_bbox', 'gt_class', 'gt_score']
+    num_max_boxes: 50
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/fruit
-    annotation: fruit-detection/train.txt
+    anno_path: train.txt
+    with_background: false
     use_default_label: false
-  num_workers: 16
-  bufsize: 128
-  use_process: true
-  mixup_epoch: -1
   sample_transforms:
   - !DecodeImage
     to_rgb: true
@@ -83,40 +83,45 @@ YoloTrainFeed:
     max_size: 0
     target_size: 608
   - !RandomFlipImage
-    is_mask_flip: false
     is_normalized: true
     prob: 0.5
   - !NormalizeImage
-    is_channel_first: false
+    mean: [0.485, 0.456, 0.406]
+    std: [0.229, 0.224, 0.225]
     is_scale: true
-    mean:
-    - 0.485
-    - 0.456
-    - 0.406
-    std:
-    - 0.229
-    - 0.224
-    - 0.225
+    is_channel_first: false
+  - !PadBox
+    num_max_boxes: 50
+  - !BboxXYXY2XYWH {}
+  batch_transforms:
+  - !RandomShape
+    sizes: [608]
   - !Permute
     channel_first: true
     to_bgr: false
-  batch_transforms:
-  - !RandomShape 
-    sizes: [608] 
-  with_background: false
+  batch_size: 1
+  shuffle: true
+  mixup_epoch: -1
+  worker_num: 4
+  bufsize: 64
+  use_process: true
 
-YoloEvalFeed:
+EvalReader:
   batch_size: 1
-  image_shape: [3, 608, 608]
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'im_size', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+    num_max_boxes: 50
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/fruit
-    annotation: fruit-detection/val.txt
+    anno_path: val.txt
     use_default_label: false
- 
+    with_background: false
 
-YoloTestFeed:
+TestReader:
   batch_size: 1
-  image_shape: [3, 608, 608]
   dataset:
-    dataset_dir: dataset/fruit
+    !ImageFolder
     use_default_label: false
+    with_background: false

configs/yolov3_mobilenet_v1_voc.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 70000
 log_smooth_window: 20
@@ -59,27 +56,29 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
+_READER_: 'yolov3_reader.yml'
+TrainReader:
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: trainval.txt
+    anno_path: trainval.txt
     use_default_label: true
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-  mixup_epoch: 250
+    with_background: false
 
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
+EvalReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'im_size', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+    num_max_boxes: 50
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: test.txt
+    anno_path: test.txt
     use_default_label: true
+    with_background: false
 
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
+TestReader:
   dataset:
+    !ImageFolder
     use_default_label: true
+    with_background: false

configs/yolov3_r34.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 500200
 log_smooth_window: 20
@@ -60,26 +57,4 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_train2017.json
-    image_dir: train2017
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
-  dataset:
-    dataset_dir: dataset/coco
-    annotation: annotations/instances_val2017.json
-    image_dir: val2017
-
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
-  dataset:
-    annotation: dataset/coco/annotations/instances_val2017.json
+_READER_: 'yolov3_reader.yml'

configs/yolov3_r34_voc.yml

 architecture: YOLOv3
-train_feed: YoloTrainFeed
-eval_feed: YoloEvalFeed
-test_feed: YoloTestFeed
 use_gpu: true
 max_iters: 70000
 log_smooth_window: 20
@@ -61,27 +58,29 @@ OptimizerBuilder:
     factor: 0.0005
     type: L2
 
-YoloTrainFeed:
-  batch_size: 8
+_READER_: 'yolov3_reader.yml'
+TrainReader:
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: trainval.txt
+    anno_path: trainval.txt
     use_default_label: true
-  num_workers: 8
-  bufsize: 128
-  use_process: true
-  mixup_epoch: 250
+    with_background: false
 
-YoloEvalFeed:
-  batch_size: 8
-  image_shape: [3, 608, 608]
+EvalReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'im_size', 'im_id', 'gt_bbox', 'gt_class', 'is_difficult']
+    num_max_boxes: 50
   dataset:
+    !VOCDataSet
     dataset_dir: dataset/voc
-    annotation: test.txt
+    anno_path: test.txt
     use_default_label: true
+    with_background: false
 
-YoloTestFeed:
-  batch_size: 1
-  image_shape: [3, 608, 608]
+TestReader:
   dataset:
+    !ImageFolder
     use_default_label: true
+    with_background: false

configs/yolov3_reader.yml

+TrainReader:
+  inputs_def:
+    fields: ['image', 'gt_bbox', 'gt_class', 'gt_score']
+    num_max_boxes: 50
+  dataset:
+    !COCODataSet
+      image_dir: train2017
+      anno_path: annotations/instances_train2017.json
+      dataset_dir: dataset/coco
+      with_background: false
+  sample_transforms:
+    - !DecodeImage
+      to_rgb: True
+      with_mixup: True
+    - !MixupImage
+      alpha: 1.5
+      beta: 1.5
+    - !ColorDistort {}
+    - !RandomExpand
+      fill_value: [123.675, 116.28, 103.53]
+    - !RandomCrop {}
+    - !RandomFlipImage
+      is_normalized: false
+    - !NormalizeBox {}
+    - !PadBox
+      num_max_boxes: 50
+    - !BboxXYXY2XYWH {}
+  batch_transforms:
+  - !RandomShape
+    sizes: [320, 352, 384, 416, 448, 480, 512, 544, 576, 608]
+    random_inter: True
+  - !NormalizeImage
+    mean: [0.485, 0.456, 0.406]
+    std: [0.229, 0.224, 0.225]
+    is_scale: True
+    is_channel_first: false
+  - !Permute
+    to_bgr: false
+    channel_first: True
+  batch_size: 8
+  shuffle: true
+  mixup_epoch: 250
+  drop_last: true
+  worker_num: 8
+  bufsize: 32
+  use_process: true
+
+
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_size', 'im_id']
+    num_max_boxes: 50
+  dataset:
+    !COCODataSet
+      image_dir: val2017
+      anno_path: annotations/instances_val2017.json
+      dataset_dir: dataset/coco
+      with_background: false
+  sample_transforms:
+    - !DecodeImage
+      to_rgb: True
+    - !ResizeImage
+      target_size: 608
+      interp: 2
+    - !NormalizeImage
+      mean: [0.485, 0.456, 0.406]
+      std: [0.229, 0.224, 0.225]
+      is_scale: True
+      is_channel_first: false
+    - !PadBox
+      num_max_boxes: 50
+    - !Permute
+      to_bgr: false
+      channel_first: True
+  batch_size: 8
+  drop_empty: false
+  worker_num: 8
+  bufsize: 32
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+    fields: ['image', 'im_size', 'im_id']
+  dataset:
+    !ImageFolder
+      anno_path: annotations/instances_val2017.json
+      with_background: false
+  sample_transforms:
+    - !DecodeImage
+      to_rgb: True
+    - !ResizeImage
+      target_size: 608
+      interp: 2
+    - !NormalizeImage
+      mean: [0.485, 0.456, 0.406]
+      std: [0.229, 0.224, 0.225]
+      is_scale: True
+      is_channel_first: false
+    - !Permute
+      to_bgr: false
+      channel_first: True
+  batch_size: 1

docs/GETTING_STARTED.md

@@ -34,7 +34,6 @@ list below can be viewed by `--help`
 |   -r/--resume_checkpoint |     train      |  Checkpoint path for resuming training  |  None  |  `-r output/faster_rcnn_r50_1x/10000`  |
 |        --eval            |     train      |  Whether to perform evaluation in training  |  False  |    |
 |      --output_eval       |     train/eval |  json path in evalution  |  current path  |  `--output_eval ./json_result`  |
-|   -d/--dataset_dir       |   train/eval   |  path for dataset, same as dataset_dir in configs  |  None  |  `-d dataset/coco`  |
 |       --fp16             |     train      |  Whether to enable mixed precision training  |  False  |  GPU training is required  |
 |       --loss_scale       |     train      |  Loss scaling factor for mixed precision training  |  8.0  |  enable when `--fp16` is True  |  
 |       --json_eval        |       eval     |  Whether to evaluate with already existed bbox.json or mask.json  |  False  |  json path is set in `--output_eval`  |

ppdet/data/README.md

-docs/DATA.md
\ No newline at end of file

ppdet/data/README_cn.md

-docs/DATA_cn.md
\ No newline at end of file