Add ppyoloe semi-det base codes (#7680)

* add ppyoloe semi-det base codes * fix configs * fix head distill loss * add more semi_det configs and fix doc, test=document_fix * add contrast_loss config, test=document_fix

Add ppyoloe semi-det base codes (#7680)
* add ppyoloe semi-det base codes * fix configs * fix head distill loss * add more semi_det configs and fix doc, test=document_fix * add contrast_loss config, test=document_fix
1e21400e · Feng Ni · GitHub · a496c2dd · 1e21400e · 1e21400e
14 changed file
--- a/configs/semi_det/baseline/README.md
+++ b/configs/semi_det/baseline/README.md
@@ -21,6 +21,9 @@
 | PP-YOLOE+_s       |        5%       |    80 (7200)      |       32.8       | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco_sup005.pdparams) | [config](ppyoloe_plus_crn_s_80e_coco_sup005.yml) |
 | PP-YOLOE+_s       |        10%      |    80 (14480)     |       35.3       | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco_sup010.pdparams) | [config](ppyoloe_plus_crn_s_80e_coco_sup010.yml) |
 | PP-YOLOE+_s       |        full     |    80 (146560)    |       43.7       | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco.pdparams) | [config](../../ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml) |
+| PP-YOLOE+_l       |        5%       |    80 (7200)      |       42.9       | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco_sup005.pdparams) | [config](ppyoloe_plus_crn_l_80e_coco_sup005.yml) |
+| PP-YOLOE+_l       |        10%      |    80 (14480)     |       45.7       | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco_sup010.pdparams) | [config](ppyoloe_plus_crn_l_80e_coco_sup010.yml) |
+| PP-YOLOE+_l       |        full     |    80 (146560)    |       49.8       | [download](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams) | [config](../../ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml) |

 **注意:**
  - 以上模型训练默认使用8 GPUs，总batch_size默认为64，默认初始学习率为0.001。如果改动了总batch_size，请按线性比例相应地调整学习率。

--- a/configs/semi_det/baseline/ppyoloe_plus_crn_l_80e_coco_sup005.yml
+++ b/configs/semi_det/baseline/ppyoloe_plus_crn_l_80e_coco_sup005.yml
+_BASE_: [
+  '../../ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml',
+]
+log_iter: 50
+snapshot_epoch: 5
+weights: output/ppyoloe_plus_crn_l_80e_coco_sup005/model_final
+
+pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/pretrained/ppyoloe_crn_l_obj365_pretrained.pdparams
+depth_mult: 1.0
+width_mult: 1.0
+
+
+TrainDataset:
+  !COCODataSet
+    image_dir: train2017
+    anno_path: semi_annotations/instances_train2017.1@5.json
+    dataset_dir: dataset/coco
+    data_fields: ['image', 'gt_bbox', 'gt_class']
+
+
+epoch: 80
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+    - !CosineDecay
+      max_epochs: 96
+    - !LinearWarmup
+      start_factor: 0.
+      epochs: 5
--- a/configs/semi_det/baseline/ppyoloe_plus_crn_l_80e_coco_sup010.yml
+++ b/configs/semi_det/baseline/ppyoloe_plus_crn_l_80e_coco_sup010.yml
+_BASE_: [
+  '../../ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml',
+]
+log_iter: 50
+snapshot_epoch: 5
+weights: output/ppyoloe_plus_crn_l_80e_coco_sup010/model_final
+
+pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/pretrained/ppyoloe_crn_l_obj365_pretrained.pdparams
+depth_mult: 1.0
+width_mult: 1.0
+
+
+TrainDataset:
+  !COCODataSet
+    image_dir: train2017
+    anno_path: semi_annotations/instances_train2017.1@10.json
+    dataset_dir: dataset/coco
+    data_fields: ['image', 'gt_bbox', 'gt_class']
+
+
+epoch: 80
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+    - !CosineDecay
+      max_epochs: 96
+    - !LinearWarmup
+      start_factor: 0.
+      epochs: 5
--- a/configs/semi_det/denseteacher/README.md
+++ b/configs/semi_det/denseteacher/README.md
@@ -2,7 +2,7 @@

 # Dense Teacher: Dense Pseudo-Labels for Semi-supervised Object Detection

-## 模型库
+## FCOS模型库

 |      模型       |  监督数据比例 |        Sup Baseline     |    Sup Epochs (Iters)   |  Sup mAP<sup>val<br>0.5:0.95 | Semi mAP<sup>val<br>0.5:0.95 |  Semi Epochs (Iters)  |  模型下载  |   配置文件   |
 | :------------: | :---------: | :---------------------: | :---------------------: |:---------------------------: |:----------------------------: | :------------------: |:--------: |:----------: |
@@ -34,6 +34,16 @@
 ```


+## PPYOLOE+ 模型库
+
+|      模型       |  监督数据比例 |        Sup Baseline     |    Sup Epochs (Iters)   |  Sup mAP<sup>val<br>0.5:0.95 | Semi mAP<sup>val<br>0.5:0.95 |  Semi Epochs (Iters)  |  模型下载  |   配置文件   |
+| :------------: | :---------: | :---------------------: | :---------------------: |:---------------------------: |:----------------------------: | :------------------: |:--------: |:----------: |
+| DenseTeacher-PPYOLOE+_s | 5% |   [sup_config](../baseline/ppyoloe_plus_crn_s_80e_coco_sup005.yml) | 80 (14480) | 32.8 |  **34.0**  | 200 (36200)  | [download](https://paddledet.bj.bcebos.com/models/denseteacher_ppyoloe_plus_crn_s_coco_semi005.pdparams) | [config](./denseteacher_ppyoloe_plus_crn_s_coco_semi005.yml) |
+| DenseTeacher-PPYOLOE+_s | 10% |   [sup_config](../baseline/ppyoloe_plus_crn_s_80e_coco_sup010.yml) | 80 (14480) | 35.3 |  **37.5**  | 200 (36200)  | [download](https://paddledet.bj.bcebos.com/models/denseteacher_ppyoloe_plus_crn_s_coco_semi010.pdparams) | [config](./denseteacher_ppyoloe_plus_crn_s_coco_semi010.yml) |
+| DenseTeacher-PPYOLOE+_l | 5% |   [sup_config](../baseline/ppyoloe_plus_crn_s_80e_coco_sup005.yml) | 80 (14480) | 42.9 |  **45.4**  | 200 (36200)  | [download](https://paddledet.bj.bcebos.com/models/denseteacher_ppyoloe_plus_crn_l_coco_semi005.pdparams) | [config](./denseteacher_ppyoloe_plus_crn_l_coco_semi005.yml) |
+| DenseTeacher-PPYOLOE+_l | 10% |   [sup_config](../baseline/ppyoloe_plus_crn_l_80e_coco_sup010.yml) | 80 (14480) | 45.7 |  **47.4**  | 200 (36200)  | [download](https://paddledet.bj.bcebos.com/models/denseteacher_ppyoloe_plus_crn_l_coco_semi010.pdparams) | [config](./denseteacher_ppyoloe_plus_crn_l_coco_semi010.yml) |
+
+
 ## 使用说明

 仅训练时必须使用半监督检测的配置文件去训练，评估、预测、部署也可以按基础检测器的配置文件去执行。

--- a/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_l_coco_semi005.yml
+++ b/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_l_coco_semi005.yml
+_BASE_: [
+  '../../ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml',
+  '../_base_/coco_detection_percent_5.yml',
+]
+log_iter: 50
+snapshot_epoch: 5
+weights: output/denseteacher_ppyoloe_plus_crn_l_coco_semi005/model_final
+
+epochs: &epochs 200
+cosine_epochs: &cosine_epochs 240
+
+
+### pretrain and warmup config, choose one and comment another
+pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/semi_det/ppyoloe_plus_crn_l_80e_coco_sup005.pdparams # mAP=42.9
+semi_start_iters: 0
+ema_start_iters: 0
+use_warmup: &use_warmup False
+
+# pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/pretrained/ppyoloe_crn_l_obj365_pretrained.pdparams
+# semi_start_iters: 5000
+# ema_start_iters: 3000
+# use_warmup: &use_warmup True
+
+
+### global config
+use_simple_ema: True
+ema_decay: 0.9996
+ssod_method: DenseTeacher
+DenseTeacher:
+  train_cfg:
+    sup_weight: 1.0
+    unsup_weight: 1.0
+    loss_weight: {distill_loss_cls: 1.0, distill_loss_iou: 2.5, distill_loss_dfl: 0., distill_loss_contrast: 0.1}
+    contrast_loss:
+      temperature: 0.2
+      alpha: 0.9
+      smooth_iter: 100
+    concat_sup_data: True
+    suppress: linear
+    ratio: 0.01
+  test_cfg:
+    inference_on: teacher
+
+
+### reader config
+batch_size: &batch_size 8
+worker_num: 2
+SemiTrainReader:
+  sample_transforms:
+    - Decode: {}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53]}
+    - RandomFlip: {}
+    - RandomCrop: {} # unsup will be fake gt_boxes
+  weak_aug:
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  strong_aug:
+    - StrongAugImage: {transforms: [
+        RandomColorJitter: {prob: 0.8, brightness: 0.4, contrast: 0.4, saturation: 0.4, hue: 0.1},
+        RandomErasingCrop: {},
+        RandomGaussianBlur: {prob: 0.5, sigma: [0.1, 2.0]},
+        RandomGrayscale: {prob: 0.2},
+      ]}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  sup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+    - PadGT: {}
+  unsup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+  sup_batch_size: *batch_size
+  unsup_batch_size: *batch_size
+  shuffle: True
+  drop_last: True
+  collate_batch: True
+
+EvalReader:
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 640, 640]
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 1
+
+
+### model config
+architecture: PPYOLOE
+norm_type: sync_bn
+ema_black_list: ['proj_conv.weight']
+custom_black_list: ['reduce_mean']
+PPYOLOE:
+  backbone: CSPResNet
+  neck: CustomCSPPAN
+  yolo_head: PPYOLOEHead
+  post_process: ~
+
+eval_size: ~ # means None, but not str 'None'
+PPYOLOEHead:
+  fpn_strides: [32, 16, 8]
+  grid_cell_scale: 5.0
+  grid_cell_offset: 0.5
+  static_assigner_epoch: -1 #
+  use_varifocal_loss: True
+  loss_weight: {class: 1.0, iou: 2.5, dfl: 0.5}
+  static_assigner:
+    name: ATSSAssigner
+    topk: 9
+  assigner:
+    name: TaskAlignedAssigner
+    topk: 13
+    alpha: 1.0
+    beta: 6.0
+  nms:
+    name: MultiClassNMS
+    nms_top_k: 1000
+    keep_top_k: 300
+    score_threshold: 0.01
+    nms_threshold: 0.7
+
+
+### other config
+epoch: *epochs
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !CosineDecay
+    max_epochs: *cosine_epochs
+    use_warmup: *use_warmup
+  - !LinearWarmup
+    start_factor: 0.001
+    epochs: 3
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0005 # dt-fcos 0.0001
+    type: L2
+  clip_grad_by_norm: 1.0 # dt-fcos clip_grad_by_value
--- a/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_l_coco_semi010.yml
+++ b/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_l_coco_semi010.yml
+_BASE_: [
+  '../../ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml',
+  '../_base_/coco_detection_percent_10.yml',
+]
+log_iter: 50
+snapshot_epoch: 5
+weights: output/denseteacher_ppyoloe_plus_crn_l_coco_semi010/model_final
+
+epochs: &epochs 200
+cosine_epochs: &cosine_epochs 240
+
+
+### pretrain and warmup config, choose one and comment another
+pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/semi_det/ppyoloe_plus_crn_l_80e_coco_sup010.pdparams # mAP=45.7
+semi_start_iters: 0
+ema_start_iters: 0
+use_warmup: &use_warmup False
+
+# pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/pretrained/ppyoloe_crn_l_obj365_pretrained.pdparams
+# semi_start_iters: 5000
+# ema_start_iters: 3000
+# use_warmup: &use_warmup True
+
+
+### global config
+use_simple_ema: True
+ema_decay: 0.9996
+ssod_method: DenseTeacher
+DenseTeacher:
+  train_cfg:
+    sup_weight: 1.0
+    unsup_weight: 1.0
+    loss_weight: {distill_loss_cls: 1.0, distill_loss_iou: 2.5, distill_loss_dfl: 0., distill_loss_contrast: 0.1}
+    contrast_loss:
+      temperature: 0.2
+      alpha: 0.9
+      smooth_iter: 100
+    concat_sup_data: True
+    suppress: linear
+    ratio: 0.01
+  test_cfg:
+    inference_on: teacher
+
+
+### reader config
+batch_size: &batch_size 8
+worker_num: 2
+SemiTrainReader:
+  sample_transforms:
+    - Decode: {}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53]}
+    - RandomFlip: {}
+    - RandomCrop: {} # unsup will be fake gt_boxes
+  weak_aug:
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  strong_aug:
+    - StrongAugImage: {transforms: [
+        RandomColorJitter: {prob: 0.8, brightness: 0.4, contrast: 0.4, saturation: 0.4, hue: 0.1},
+        RandomErasingCrop: {},
+        RandomGaussianBlur: {prob: 0.5, sigma: [0.1, 2.0]},
+        RandomGrayscale: {prob: 0.2},
+      ]}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  sup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+    - PadGT: {}
+  unsup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+  sup_batch_size: *batch_size
+  unsup_batch_size: *batch_size
+  shuffle: True
+  drop_last: True
+  collate_batch: True
+
+EvalReader:
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 640, 640]
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 1
+
+
+### model config
+architecture: PPYOLOE
+norm_type: sync_bn
+ema_black_list: ['proj_conv.weight']
+custom_black_list: ['reduce_mean']
+PPYOLOE:
+  backbone: CSPResNet
+  neck: CustomCSPPAN
+  yolo_head: PPYOLOEHead
+  post_process: ~
+
+eval_size: ~ # means None, but not str 'None'
+PPYOLOEHead:
+  fpn_strides: [32, 16, 8]
+  grid_cell_scale: 5.0
+  grid_cell_offset: 0.5
+  static_assigner_epoch: -1 #
+  use_varifocal_loss: True
+  loss_weight: {class: 1.0, iou: 2.5, dfl: 0.5}
+  static_assigner:
+    name: ATSSAssigner
+    topk: 9
+  assigner:
+    name: TaskAlignedAssigner
+    topk: 13
+    alpha: 1.0
+    beta: 6.0
+  nms:
+    name: MultiClassNMS
+    nms_top_k: 1000
+    keep_top_k: 300
+    score_threshold: 0.01
+    nms_threshold: 0.7
+
+
+### other config
+epoch: *epochs
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !CosineDecay
+    max_epochs: *cosine_epochs
+    use_warmup: *use_warmup
+  - !LinearWarmup
+    start_factor: 0.001
+    epochs: 3
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0005 # dt-fcos 0.0001
+    type: L2
+  clip_grad_by_norm: 1.0 # dt-fcos clip_grad_by_value
--- a/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_s_coco_semi005.yml
+++ b/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_s_coco_semi005.yml
+_BASE_: [
+  '../../ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml',
+  '../_base_/coco_detection_percent_5.yml',
+]
+log_iter: 50
+snapshot_epoch: 5
+weights: output/denseteacher_ppyoloe_plus_crn_s_coco_semi005/model_final
+
+epochs: &epochs 200
+cosine_epochs: &cosine_epochs 240
+
+
+### pretrain and warmup config, choose one and comment another
+pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/semi_det/ppyoloe_plus_crn_s_80e_coco_sup005.pdparams # mAP=32.8
+semi_start_iters: 0
+ema_start_iters: 0
+use_warmup: &use_warmup False
+
+# pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/pretrained/ppyoloe_crn_s_obj365_pretrained.pdparams
+# semi_start_iters: 5000
+# ema_start_iters: 3000
+# use_warmup: &use_warmup True
+
+
+### global config
+use_simple_ema: True
+ema_decay: 0.9996
+ssod_method: DenseTeacher
+DenseTeacher:
+  train_cfg:
+    sup_weight: 1.0
+    unsup_weight: 1.0
+    loss_weight: {distill_loss_cls: 1.0, distill_loss_iou: 2.5, distill_loss_dfl: 0., distill_loss_contrast: 0.1}
+    contrast_loss:
+      temperature: 0.2
+      alpha: 0.9
+      smooth_iter: 100
+    concat_sup_data: True
+    suppress: linear
+    ratio: 0.01
+  test_cfg:
+    inference_on: teacher
+
+
+### reader config
+batch_size: &batch_size 8
+worker_num: 2
+SemiTrainReader:
+  sample_transforms:
+    - Decode: {}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53]}
+    - RandomFlip: {}
+    - RandomCrop: {} # unsup will be fake gt_boxes
+  weak_aug:
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  strong_aug:
+    - StrongAugImage: {transforms: [
+        RandomColorJitter: {prob: 0.8, brightness: 0.4, contrast: 0.4, saturation: 0.4, hue: 0.1},
+        RandomErasingCrop: {},
+        RandomGaussianBlur: {prob: 0.5, sigma: [0.1, 2.0]},
+        RandomGrayscale: {prob: 0.2},
+      ]}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  sup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+    - PadGT: {}
+  unsup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+  sup_batch_size: *batch_size
+  unsup_batch_size: *batch_size
+  shuffle: True
+  drop_last: True
+  collate_batch: True
+
+EvalReader:
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 640, 640]
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 1
+
+
+### model config
+architecture: PPYOLOE
+norm_type: sync_bn
+ema_black_list: ['proj_conv.weight']
+custom_black_list: ['reduce_mean']
+PPYOLOE:
+  backbone: CSPResNet
+  neck: CustomCSPPAN
+  yolo_head: PPYOLOEHead
+  post_process: ~
+
+eval_size: ~ # means None, but not str 'None'
+PPYOLOEHead:
+  fpn_strides: [32, 16, 8]
+  grid_cell_scale: 5.0
+  grid_cell_offset: 0.5
+  static_assigner_epoch: -1 #
+  use_varifocal_loss: True
+  loss_weight: {class: 1.0, iou: 2.5, dfl: 0.5}
+  static_assigner:
+    name: ATSSAssigner
+    topk: 9
+  assigner:
+    name: TaskAlignedAssigner
+    topk: 13
+    alpha: 1.0
+    beta: 6.0
+  nms:
+    name: MultiClassNMS
+    nms_top_k: 1000
+    keep_top_k: 300
+    score_threshold: 0.01
+    nms_threshold: 0.7
+
+
+### other config
+epoch: *epochs
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !CosineDecay
+    max_epochs: *cosine_epochs
+    use_warmup: *use_warmup
+  - !LinearWarmup
+    start_factor: 0.001
+    epochs: 3
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0005 # dt-fcos 0.0001
+    type: L2
+  clip_grad_by_norm: 1.0 # dt-fcos clip_grad_by_value
--- a/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_s_coco_semi010.yml
+++ b/configs/semi_det/denseteacher/denseteacher_ppyoloe_plus_crn_s_coco_semi010.yml
+_BASE_: [
+  '../../ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml',
+  '../_base_/coco_detection_percent_10.yml',
+]
+log_iter: 50
+snapshot_epoch: 5
+weights: output/denseteacher_ppyoloe_plus_crn_s_coco_semi010/model_final
+
+epochs: &epochs 200
+cosine_epochs: &cosine_epochs 240
+
+
+### pretrain and warmup config, choose one and comment another
+pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/semi_det/ppyoloe_plus_crn_s_80e_coco_sup010.pdparams # mAP=35.3
+semi_start_iters: 0
+ema_start_iters: 0
+use_warmup: &use_warmup False
+
+# pretrain_weights: https://bj.bcebos.com/v1/paddledet/models/pretrained/ppyoloe_crn_s_obj365_pretrained.pdparams
+# semi_start_iters: 5000
+# ema_start_iters: 3000
+# use_warmup: &use_warmup True
+
+
+### global config
+use_simple_ema: True
+ema_decay: 0.9996
+ssod_method: DenseTeacher
+DenseTeacher:
+  train_cfg:
+    sup_weight: 1.0
+    unsup_weight: 1.0
+    loss_weight: {distill_loss_cls: 1.0, distill_loss_iou: 2.5, distill_loss_dfl: 0., distill_loss_contrast: 0.1}
+    contrast_loss:
+      temperature: 0.2
+      alpha: 0.9
+      smooth_iter: 100
+    concat_sup_data: True
+    suppress: linear
+    ratio: 0.01
+  test_cfg:
+    inference_on: teacher
+
+
+### reader config
+batch_size: &batch_size 8
+worker_num: 2
+SemiTrainReader:
+  sample_transforms:
+    - Decode: {}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53]}
+    - RandomFlip: {}
+    - RandomCrop: {} # unsup will be fake gt_boxes
+  weak_aug:
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  strong_aug:
+    - StrongAugImage: {transforms: [
+        RandomColorJitter: {prob: 0.8, brightness: 0.4, contrast: 0.4, saturation: 0.4, hue: 0.1},
+        RandomErasingCrop: {},
+        RandomGaussianBlur: {prob: 0.5, sigma: [0.1, 2.0]},
+        RandomGrayscale: {prob: 0.2},
+      ]}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], is_scale: true, norm_type: none}
+  sup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+    - PadGT: {}
+  unsup_batch_transforms:
+    - BatchRandomResize: {target_size: [640], random_size: True, random_interp: True, keep_ratio: False}
+    - Permute: {}
+  sup_batch_size: *batch_size
+  unsup_batch_size: *batch_size
+  shuffle: True
+  drop_last: True
+  collate_batch: True
+
+EvalReader:
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 2
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 640, 640]
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [640, 640], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0., 0., 0.], std: [1., 1., 1.], norm_type: none}
+    - Permute: {}
+  batch_size: 1
+
+
+### model config
+architecture: PPYOLOE
+norm_type: sync_bn
+ema_black_list: ['proj_conv.weight']
+custom_black_list: ['reduce_mean']
+PPYOLOE:
+  backbone: CSPResNet
+  neck: CustomCSPPAN
+  yolo_head: PPYOLOEHead
+  post_process: ~
+
+eval_size: ~ # means None, but not str 'None'
+PPYOLOEHead:
+  fpn_strides: [32, 16, 8]
+  grid_cell_scale: 5.0
+  grid_cell_offset: 0.5
+  static_assigner_epoch: -1 #
+  use_varifocal_loss: True
+  loss_weight: {class: 1.0, iou: 2.5, dfl: 0.5}
+  static_assigner:
+    name: ATSSAssigner
+    topk: 9
+  assigner:
+    name: TaskAlignedAssigner
+    topk: 13
+    alpha: 1.0
+    beta: 6.0
+  nms:
+    name: MultiClassNMS
+    nms_top_k: 1000
+    keep_top_k: 300
+    score_threshold: 0.01
+    nms_threshold: 0.7
+
+
+### other config
+epoch: *epochs
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !CosineDecay
+    max_epochs: *cosine_epochs
+    use_warmup: *use_warmup
+  - !LinearWarmup
+    start_factor: 0.001
+    epochs: 3
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0005 # dt-fcos 0.0001
+    type: L2
+  clip_grad_by_norm: 1.0 # dt-fcos clip_grad_by_value
--- a/ppdet/data/transform/operators.py
+++ b/ppdet/data/transform/operators.py
@@ -1420,10 +1420,38 @@ class RandomCrop(BaseOperator):
                crop_segms.append(_crop_rle(segm, crop, height, width))
        return crop_segms

+    def set_fake_bboxes(self, sample):
+        sample['gt_bbox'] = np.array(
+            [
+                [32, 32, 128, 128],
+                [32, 32, 128, 256],
+                [32, 64, 128, 128],
+                [32, 64, 128, 256],
+                [64, 64, 128, 256],
+                [64, 64, 256, 256],
+                [64, 32, 128, 256],
+                [64, 32, 128, 256],
+                [96, 32, 128, 256],
+                [96, 32, 128, 256],
+            ],
+            dtype=np.float32)
+        sample['gt_class'] = np.array(
+            [[1], [2], [3], [4], [5], [6], [7], [8], [9], [10]], np.int32)
+        return sample
+
    def apply(self, sample, context=None):
+        if 'gt_bbox' not in sample:
+            # only used in semi-det as unsup data
+            sample = self.set_fake_bboxes(sample)
+            sample = self.random_crop(sample, fake_bboxes=True)
+            return sample
+
        if 'gt_bbox' in sample and len(sample['gt_bbox']) == 0:
            return sample
+        sample = self.random_crop(sample)
+        return sample

+    def random_crop(self, sample, fake_bboxes=False):
        h, w = sample['image'].shape[:2]
        gt_bbox = sample['gt_bbox']

@@ -1515,6 +1543,9 @@ class RandomCrop(BaseOperator):
                        sample['gt_segm'], valid_ids, axis=0)

                sample['image'] = self._crop_image(sample['image'], crop_box)
+                if fake_bboxes == True:
+                    return sample
+
                sample['gt_bbox'] = np.take(cropped_box, valid_ids, axis=0)
                sample['gt_class'] = np.take(
                    sample['gt_class'], valid_ids, axis=0)

--- a/ppdet/engine/trainer_ssod.py
+++ b/ppdet/engine/trainer_ssod.py
@@ -16,12 +16,9 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function

-import os
-import sys
 import copy
 import time
 import typing
-import math
 import numpy as np

 import paddle
@@ -317,16 +314,14 @@ class Trainer_DenseTeacher(Trainer):
                        data_unsup_w['is_teacher'] = True
                        teacher_preds = self.ema.model(data_unsup_w)

+                    train_cfg['curr_iter'] = curr_iter
+                    train_cfg['st_iter'] = st_iter
                    if self._nranks > 1:
-                        loss_dict_unsup = self.model._layers.get_distill_loss(
-                            student_preds,
-                            teacher_preds,
-                            ratio=train_cfg['ratio'])
+                        loss_dict_unsup = self.model._layers.get_ssod_distill_loss(
+                            student_preds, teacher_preds, train_cfg)
                    else:
-                        loss_dict_unsup = self.model.get_distill_loss(
-                            student_preds,
-                            teacher_preds,
-                            ratio=train_cfg['ratio'])
+                        loss_dict_unsup = self.model.get_ssod_distill_loss(
+                            student_preds, teacher_preds, train_cfg)

                    fg_num = loss_dict_unsup["fg_sum"]
                    del loss_dict_unsup["fg_sum"]

--- a/ppdet/modeling/architectures/fcos.py
+++ b/ppdet/modeling/architectures/fcos.py
@@ -85,12 +85,10 @@ class FCOS(BaseArch):
    def get_loss_keys(self):
        return ['loss_cls', 'loss_box', 'loss_quality']

-    def get_distill_loss(self,
-                         fcos_head_outs,
-                         teacher_fcos_head_outs,
-                         ratio=0.01):
-        student_logits, student_deltas, student_quality = fcos_head_outs
-        teacher_logits, teacher_deltas, teacher_quality = teacher_fcos_head_outs
+    def get_ssod_distill_loss(self, student_head_outs, teacher_head_outs,
+                              train_cfg):
+        student_logits, student_deltas, student_quality = student_head_outs
+        teacher_logits, teacher_deltas, teacher_quality = teacher_head_outs
        nc = student_logits[0].shape[1]

        student_logits = paddle.concat(
@@ -132,6 +130,7 @@ class FCOS(BaseArch):
            ],
            axis=0)

+        ratio = train_cfg.get('ratio', 0.01)
        with paddle.no_grad():
            # Region Selection
            count_num = int(teacher_logits.shape[0] * ratio)

--- a/ppdet/modeling/architectures/ppyoloe.py
+++ b/ppdet/modeling/architectures/ppyoloe.py
@@ -16,10 +16,14 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function

-import paddle
 import copy
+
+import paddle
+import paddle.nn.functional as F
 from ppdet.core.workspace import register, create
 from .meta_arch import BaseArch
+from ..ssod_utils import QFLv2
+from ..losses import GIoULoss

 __all__ = ['PPYOLOE', 'PPYOLOEWithAuxHead']
 # PP-YOLOE and PP-YOLOE+ are recommended to use this architecture, especially when use distillation or aux head
@@ -57,6 +61,11 @@ class PPYOLOE(BaseArch):
        self.yolo_head = yolo_head
        self.post_process = post_process
        self.for_mot = for_mot
+
+        # semi-det
+        self.is_teacher = False
+
+        # distill
        self.for_distill = for_distill
        self.feat_distill_place = feat_distill_place
        if for_distill:
@@ -85,7 +94,8 @@ class PPYOLOE(BaseArch):
        body_feats = self.backbone(self.inputs)
        neck_feats = self.neck(body_feats, self.for_mot)

-        if self.training:
+        self.is_teacher = self.inputs.get('is_teacher', False)  # for semi-det
+        if self.training or self.is_teacher:
            yolo_losses = self.yolo_head(neck_feats, self.inputs)

            if self.for_distill:
@@ -121,6 +131,110 @@ class PPYOLOE(BaseArch):
    def get_pred(self):
        return self._forward()

+    def get_loss_keys(self):
+        return ['loss_cls', 'loss_iou', 'loss_dfl', 'loss_contrast']
+
+    def get_ssod_distill_loss(self, student_head_outs, teacher_head_outs,
+                              train_cfg):
+        # for semi-det distill
+        # student_probs: already sigmoid
+        student_probs, student_deltas, student_dfl = student_head_outs
+        teacher_probs, teacher_deltas, teacher_dfl = teacher_head_outs
+        bs, l, nc = student_probs.shape[:]
+        student_probs = student_probs.reshape([-1, nc])
+        teacher_probs = teacher_probs.reshape([-1, nc])
+        student_deltas = student_deltas.reshape([-1, 4])
+        teacher_deltas = teacher_deltas.reshape([-1, 4])
+        student_dfl = student_dfl.reshape([-1, 4, self.yolo_head.reg_channels])
+        teacher_dfl = teacher_dfl.reshape([-1, 4, self.yolo_head.reg_channels])
+
+        ratio = train_cfg.get('ratio', 0.01)
+
+        # for contrast loss
+        curr_iter = train_cfg['curr_iter']
+        st_iter = train_cfg['st_iter']
+        if curr_iter == st_iter + 1:
+            # start semi-det training
+            self.queue_ptr = 0
+            self.queue_size = int(bs * l * ratio)
+            self.queue_feats = paddle.zeros([self.queue_size, nc])
+            self.queue_probs = paddle.zeros([self.queue_size, nc])
+        contrast_loss_cfg = train_cfg['contrast_loss']
+        temperature = contrast_loss_cfg.get('temperature', 0.2)
+        alpha = contrast_loss_cfg.get('alpha', 0.9)
+        smooth_iter = contrast_loss_cfg.get('smooth_iter', 100) + st_iter
+
+        with paddle.no_grad():
+            # Region Selection
+            count_num = int(teacher_probs.shape[0] * ratio)
+            max_vals = paddle.max(teacher_probs, 1)
+            sorted_vals, sorted_inds = paddle.topk(max_vals,
+                                                   teacher_probs.shape[0])
+            mask = paddle.zeros_like(max_vals)
+            mask[sorted_inds[:count_num]] = 1.
+            fg_num = sorted_vals[:count_num].sum()
+            b_mask = mask > 0.
+
+            # for contrast loss
+            probs = teacher_probs[b_mask].detach()
+            if curr_iter > smooth_iter:  # memory-smoothing
+                A = paddle.exp(
+                    paddle.mm(teacher_probs[b_mask], self.queue_probs.t()) /
+                    temperature)
+                A = A / A.sum(1, keepdim=True)
+                probs = alpha * probs + (1 - alpha) * paddle.mm(
+                    A, self.queue_probs)
+            n = student_probs[b_mask].shape[0]
+            # update memory bank
+            self.queue_feats[self.queue_ptr:self.queue_ptr +
+                             n, :] = teacher_probs[b_mask].detach()
+            self.queue_probs[self.queue_ptr:self.queue_ptr +
+                             n, :] = teacher_probs[b_mask].detach()
+            self.queue_ptr = (self.queue_ptr + n) % self.queue_size
+
+        # embedding similarity
+        sim = paddle.exp(
+            paddle.mm(student_probs[b_mask], teacher_probs[b_mask].t()) / 0.2)
+        sim_probs = sim / sim.sum(1, keepdim=True)
+        # pseudo-label graph with self-loop
+        Q = paddle.mm(probs, probs.t())
+        Q.fill_diagonal_(1)
+        pos_mask = (Q >= 0.5).astype('float32')
+        Q = Q * pos_mask
+        Q = Q / Q.sum(1, keepdim=True)
+        # contrastive loss
+        loss_contrast = -(paddle.log(sim_probs + 1e-7) * Q).sum(1)
+        loss_contrast = loss_contrast.mean()
+
+        # distill_loss_cls
+        loss_cls = QFLv2(
+            student_probs, teacher_probs, weight=mask, reduction="sum") / fg_num
+
+        # distill_loss_iou
+        inputs = paddle.concat(
+            (-student_deltas[b_mask][..., :2], student_deltas[b_mask][..., 2:]),
+            -1)
+        targets = paddle.concat(
+            (-teacher_deltas[b_mask][..., :2], teacher_deltas[b_mask][..., 2:]),
+            -1)
+        iou_loss = GIoULoss(reduction='mean')
+        loss_iou = iou_loss(inputs, targets)
+
+        # distill_loss_dfl
+        loss_dfl = F.cross_entropy(
+            student_dfl[b_mask].reshape([-1, self.yolo_head.reg_channels]),
+            teacher_dfl[b_mask].reshape([-1, self.yolo_head.reg_channels]),
+            soft_label=True,
+            reduction='mean')
+
+        return {
+            "distill_loss_cls": loss_cls,
+            "distill_loss_iou": loss_iou,
+            "distill_loss_dfl": loss_dfl,
+            "distill_loss_contrast": loss_contrast,
+            "fg_sum": fg_num,
+        }
+

 @register
 class PPYOLOEWithAuxHead(BaseArch):

--- a/ppdet/modeling/heads/ppyoloe_head.py
+++ b/ppdet/modeling/heads/ppyoloe_head.py
@@ -112,6 +112,7 @@ class PPYOLOEHead(nn.Layer):
        self.exclude_post_process = exclude_post_process
        self.use_shared_conv = use_shared_conv
        self.for_distill = for_distill
+        self.is_teacher = False

        # stem
        self.stem_cls = nn.LayerList()
@@ -181,6 +182,14 @@ class PPYOLOEHead(nn.Layer):
        cls_score_list = paddle.concat(cls_score_list, axis=1)
        reg_distri_list = paddle.concat(reg_distri_list, axis=1)

+        if targets.get('is_teacher', False):
+            pred_deltas, pred_dfls = self._bbox_decode_fake(reg_distri_list)
+            return cls_score_list, pred_deltas * stride_tensor, pred_dfls
+
+        if targets.get('get_data', False):
+            pred_deltas, pred_dfls = self._bbox_decode_fake(reg_distri_list)
+            return cls_score_list, pred_deltas * stride_tensor, pred_dfls
+
        return self.get_loss([
            cls_score_list, reg_distri_list, anchors, anchor_points,
            num_anchors_list, stride_tensor
@@ -249,6 +258,14 @@ class PPYOLOEHead(nn.Layer):
        if self.training:
            return self.forward_train(feats, targets, aux_pred)
        else:
+            if targets is not None:
+                # only for semi-det
+                self.is_teacher = targets.get('is_teacher', False)
+                if self.is_teacher:
+                    return self.forward_train(feats, targets, aux_pred=None)
+                else:
+                    return self.forward_eval(feats)
+
            return self.forward_eval(feats)

    @staticmethod
@@ -274,6 +291,14 @@ class PPYOLOEHead(nn.Layer):
        pred_dist = self.proj_conv(pred_dist.transpose([0, 3, 1, 2])).squeeze(1)
        return batch_distance2bbox(anchor_points, pred_dist)

+    def _bbox_decode_fake(self, pred_dist):
+        _, l, _ = get_static_shape(pred_dist)
+        pred_dist_dfl = F.softmax(
+            pred_dist.reshape([-1, l, 4, self.reg_channels]))
+        pred_dist = self.proj_conv(pred_dist_dfl.transpose([0, 3, 1, 2
+                                                            ])).squeeze(1)
+        return pred_dist, pred_dist_dfl
+
    def _bbox2distance(self, points, bbox):
        x1y1, x2y2 = paddle.split(bbox, 2, -1)
        lt = points - x1y1
@@ -388,11 +413,13 @@ class PPYOLOEHead(nn.Layer):
                            gt_bboxes,
                            pad_gt_mask,
                            bg_index=self.num_classes)
-                        self.assigned_labels = assigned_labels
-                        self.assigned_bboxes = assigned_bboxes
-                        self.assigned_scores = assigned_scores
-                        self.mask_positive = mask_positive
+                        if self.for_distill:
+                            self.assigned_labels = assigned_labels
+                            self.assigned_bboxes = assigned_bboxes
+                            self.assigned_scores = assigned_scores
+                            self.mask_positive = mask_positive
                    else:
+                        # only used in distill
                        assigned_labels = self.assigned_labels
                        assigned_bboxes = self.assigned_bboxes
                        assigned_scores = self.assigned_scores

--- a/ppdet/modeling/ssod_utils.py
+++ b/ppdet/modeling/ssod_utils.py
@@ -35,12 +35,12 @@ def align_weak_strong_shape(data_weak, data_strong):
            mode='bilinear',
            align_corners=False)
        if 'gt_bbox' in data_strong:
-            gt_bboxes = data_strong['gt_bbox']
+            gt_bboxes = data_strong['gt_bbox'].numpy()
            for i in range(len(gt_bboxes)):
                if len(gt_bboxes[i]) > 0:
                    gt_bboxes[i][:, 0::2] = gt_bboxes[i][:, 0::2] * scale_x_s
                    gt_bboxes[i][:, 1::2] = gt_bboxes[i][:, 1::2] * scale_y_s
-            data_strong['gt_bbox'] = gt_bboxes
+            data_strong['gt_bbox'] = paddle.to_tensor(gt_bboxes)

    if scale_x_w != 1 or scale_y_w != 1:
        data_weak['image'] = F.interpolate(
@@ -49,12 +49,12 @@ def align_weak_strong_shape(data_weak, data_strong):
            mode='bilinear',
            align_corners=False)
        if 'gt_bbox' in data_weak:
-            gt_bboxes = data_weak['gt_bbox']
+            gt_bboxes = data_weak['gt_bbox'].numpy()
            for i in range(len(gt_bboxes)):
                if len(gt_bboxes[i]) > 0:
                    gt_bboxes[i][:, 0::2] = gt_bboxes[i][:, 0::2] * scale_x_w
                    gt_bboxes[i][:, 1::2] = gt_bboxes[i][:, 1::2] * scale_y_w
-            data_weak['gt_bbox'] = gt_bboxes
+            data_weak['gt_bbox'] = paddle.to_tensor(gt_bboxes)
    return data_weak, data_strong