add retinanet_r101 (#2581)

* test=develop add retinanet_r101 and delete needless configures in data feeder

* test=develop
add retinanet_r101 and delete needless configures in data feeder

PaddleCV/object_detection/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
+weights: output/retinanet_r101_fpn_1x/model_final
+log_smooth_window: 20
+snapshot_iter: 10000
+metric: COCO
+save_dir: output
+
+RetinaNet:
+  backbone: ResNet
+  fpn: FPN
+  retina_head: RetinaHead
+
+ResNet:
+  norm_type: affine_channel
+  norm_decay: 0.
+  depth: 101
+  feature_maps: [3, 4, 5]
+  freeze_at: 2
+
+FPN:
+  max_level: 7
+  min_level: 3
+  num_chan: 256
+  spatial_scale: [0.03125, 0.0625, 0.125]
+  has_extra_convs: true
+
+RetinaHead:
+  num_convs_per_octave: 4
+  num_chan: 256
+  max_level: 7
+  min_level: 3
+  prior_prob: 0.01
+  base_scale: 4
+  num_scales_per_octave: 3
+  num_classes: 81
+  anchor_generator:
+    aspect_ratios: [1.0, 2.0, 0.5]
+    variance: [1.0, 1.0, 1.0, 1.0]
+  target_assign:
+    positive_overlap: 0.5
+    negative_overlap: 0.4
+  gamma: 2.0
+  alpha: 0.25
+  sigma: 3.0151134457776365
+  output_decoder:
+    score_thresh: 0.05
+    nms_thresh: 0.5
+    pre_nms_top_n: 1000
+    detections_per_im: 100
+    nms_eta: 1.0
+
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [60000, 80000]
+    values: null
+  - !LinearWarmup
+    start_factor: 0.3333333333333333
+    steps: 500
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0001
+    type: L2
+
+FasterRCNNTrainFeed:
+  batch_size: 2
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 128
+  dataset:
+    dataset_dir: data/coco
+    annotation: annotations/instances_train2017.json
+    image_dir: train2017
+  num_workers: 2
+  shuffle: True
+
+FasterRCNNEvalFeed:
+  batch_size: 2
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 128
+  dataset:
+    dataset_dir: data/coco
+    annotation: annotations/instances_val2017.json
+    image_dir: val2017
+  num_workers: 2
+
+FasterRCNNTestFeed:
+  batch_size: 1
+  batch_transforms:
+  - !PadBatch
+    pad_to_stride: 128
+  dataset:
+    annotation: annotations/instances_val2017.json
+  num_workers: 2
+  shuffle: False

PaddleCV/object_detection/configs/retinanet_r50_fpn_1x.yml

@@ -3,7 +3,7 @@ train_feed: FasterRCNNTrainFeed
 eval_feed: FasterRCNNEvalFeed
 test_feed: FasterRCNNTestFeed
 max_iters: 90000
-use_gpu: yes
+use_gpu: True
 pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_pretrained.tar
 weights: output/retinanet_r50_fpn_1x/model_final
 log_smooth_window: 20
@@ -18,12 +18,10 @@ RetinaNet:
 
 ResNet:
   norm_type: affine_channel
-  freeze_norm: true
-  norm_decay: 0.0001
+  norm_decay: 0.
   depth: 50
   feature_maps: [3, 4, 5]
   freeze_at: 2
-  variant: b
 
 FPN:
   max_level: 7
@@ -85,31 +83,8 @@ FasterRCNNTrainFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_train2017.json
     image_dir: train2017
-  drop_last: false
-  image_shape: [3, 1333, 800]
   num_workers: 2
-  sample_transforms:
-  - !DecodeImage
-    to_rgb: true
-    with_mixup: false
-  - !RandomFlipImage
-    is_mask_flip: false
-    is_normalized: false
-    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
-    interp: 1
-    max_size: 1333
-    target_size: 800
-    use_cv2: true
-  - !Permute
-    channel_first: true
-    to_bgr: false
-  shuffle: true
+  shuffle: True
 
 FasterRCNNEvalFeed:
   batch_size: 2
@@ -120,27 +95,7 @@ FasterRCNNEvalFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  shuffle: false
-  image_shape: [3, 1333, 800]
   num_workers: 2
-  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
-  shuffle: false
 
 FasterRCNNTestFeed:
   batch_size: 1
@@ -149,24 +104,5 @@ FasterRCNNTestFeed:
     pad_to_stride: 128
   dataset:
     annotation: annotations/instances_val2017.json
-  drop_last: false
-  image_shape: [3, 1333, 800]
   num_workers: 2
-  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
-  shuffle: false
+  shuffle: False