Swin faster rcnn v1 (#4050)

* add faster rcnn with swin_transformer tiny

configs/faster_rcnn/_base_/faster_rcnn_swin_transformer.yml

+architecture: FasterRCNN
+
+FasterRCNN:
+  backbone: SwinTransformer
+  neck: FPN
+  rpn_head: RPNHead
+  bbox_head: BBoxHead
+  bbox_post_process: BBoxPostProcess
+
+SwinTransformer:
+  embed_dim: 96
+  depths: [2, 2, 6, 2]
+  num_heads: [3, 6, 12, 24]
+  window_size: 7
+  ape: false
+  drop_path_rate: 0.1
+  patch_norm: true
+  out_indices: [0,1,2,3]
+  drop_path_rate: 0.1
+  pretrained: https://paddledet.bj.bcebos.com/models/pretrained/swin_tiny_patch4_window7_224.pdparams
+
+FPN:
+  out_channel: 256
+
+RPNHead:
+  anchor_generator:
+    aspect_ratios: [0.5, 1.0, 2.0]
+    anchor_sizes: [[32], [64], [128], [256], [512]]
+    strides: [4, 8, 16, 32, 64]
+  rpn_target_assign:
+    batch_size_per_im: 256
+    fg_fraction: 0.5
+    negative_overlap: 0.3
+    positive_overlap: 0.7
+    use_random: True
+  train_proposal:
+    min_size: 0.0
+    nms_thresh: 0.7
+    pre_nms_top_n: 2000
+    post_nms_top_n: 1000
+    topk_after_collect: True
+  test_proposal:
+    min_size: 0.0
+    nms_thresh: 0.7
+    pre_nms_top_n: 1000
+    post_nms_top_n: 1000
+
+
+BBoxHead:
+  head: TwoFCHead
+  roi_extractor:
+    resolution: 7
+    sampling_ratio: 0
+    aligned: True
+  bbox_assigner: BBoxAssigner
+
+BBoxAssigner:
+  batch_size_per_im: 512
+  bg_thresh: 0.5
+  fg_thresh: 0.5
+  fg_fraction: 0.25
+  use_random: True
+
+TwoFCHead:
+  out_channel: 1024
+
+BBoxPostProcess:
+  decode: RCNNBox
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.05
+    nms_threshold: 0.5

configs/faster_rcnn/_base_/faster_rcnn_swin_transformer_reader.yml

+worker_num: 2
+TrainReader:
+  sample_transforms:
+  - Decode: {}
+  - RandomResizeCrop: {resizes: [400, 500, 600], cropsizes: [[384, 600], ], prob: 0.5}
+  - RandomResize: {target_size: [[480, 1333], [512, 1333], [544, 1333], [576, 1333], [608, 1333], [640, 1333], [672, 1333], [704, 1333], [736, 1333], [768, 1333], [800, 1333]], keep_ratio: True, interp: 2}
+  - RandomFlip: {prob: 0.5}
+  - NormalizeImage: {is_scale: true, mean: [0.485,0.456,0.406], std: [0.229, 0.224,0.225]}
+  - Permute: {}
+  batch_transforms:
+  - PadBatch: {pad_to_stride: 32}
+  batch_size: 2
+  shuffle: true
+  drop_last: true
+  collate_batch: false
+
+EvalReader:
+  sample_transforms:
+  - Decode: {}
+  - Resize: {interp: 2, target_size: [800, 1333], keep_ratio: True}
+  - NormalizeImage: {is_scale: true, mean: [0.485,0.456,0.406], std: [0.229, 0.224,0.225]}
+  - Permute: {}
+  batch_transforms:
+  - PadBatch: {pad_to_stride: 32}
+  batch_size: 1
+  shuffle: false
+  drop_last: false
+  drop_empty: false
+
+
+TestReader:
+  sample_transforms:
+  - Decode: {}
+  - Resize: {interp: 2, target_size: [800, 1333], keep_ratio: True}
+  - NormalizeImage: {is_scale: true, mean: [0.485,0.456,0.406], std: [0.229, 0.224,0.225]}
+  - Permute: {}
+  batch_transforms:
+  - PadBatch: {pad_to_stride: 32}
+  batch_size: 1
+  shuffle: false
+  drop_last: false

configs/faster_rcnn/_base_/optimizer_swin_transformer_1x.yml

+epoch: 12
+
+LearningRate:
+  base_lr: 0.0001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [8, 11]
+  - !LinearWarmup
+    start_factor: 0.1
+    steps: 1000
+
+OptimizerBuilder:
+  clip_grad_by_norm: 1.0
+  optimizer:
+    type: AdamW
+    weight_decay: 0.05

configs/faster_rcnn/faster_rcnn_swin_transformer_tiny_1x_coco.yml

+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_swin_transformer_1x.yml',
+  '_base_/faster_rcnn_swin_transformer.yml',
+  '_base_/faster_rcnn_swin_transformer_reader.yml',
+]
+weights: output/faster_swin_transformer_tiny_1x/model_final

configs/faster_rcnn/faster_rcnn_swin_transformer_tiny_2x_coco.yml

+_BASE_: [
+  'faster_rcnn_swin_transformer_tiny_1x_coco.yml',
+]
+weights: output/faster_swin_transformer_tiny_2x/model_final
+
+epoch: 24
+
+LearningRate:
+  base_lr: 0.0001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [16, 22]
+  - !LinearWarmup
+    start_factor: 0.1
+    steps: 1000
+
+OptimizerBuilder:
+  clip_grad_by_norm: 1.0
+  optimizer:
+    type: AdamW
+    weight_decay: 0.05

configs/faster_rcnn/faster_rcnn_swin_transformer_tiny_3x_coco.yml

+_BASE_: [
+  'faster_rcnn_swin_transformer_tiny_1x_coco.yml',
+]
+weights: output/faster_swin_transformer_tiny_3x/model_final
+
+epoch: 36
+
+LearningRate:
+  base_lr: 0.0001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [24, 33]
+  - !LinearWarmup
+    start_factor: 0.1
+    steps: 1000
+
+OptimizerBuilder:
+  clip_grad_by_norm: 1.0
+  optimizer:
+    type: AdamW
+    weight_decay: 0.05

ppdet/modeling/backbones/__init__.py

@@ -25,6 +25,7 @@ from . import senet
 from . import res2net
 from . import dla
 from . import shufflenet_v2
+from . import swin_transformer
 
 from .vgg import *
 from .resnet import *
@@ -39,3 +40,4 @@ from .senet import *
 from .res2net import *
 from .dla import *
 from .shufflenet_v2 import *
+from .swin_transformer import *