更换文档检测模型

2024-08-27 14:42:45 +08:00
parent aea6f19951
commit 1514e09c40
2072 changed files with 254336 additions and 4967 deletions
--- a/paddle_detection/configs/yolov3/README.md
+++ b/paddle_detection/configs/yolov3/README.md
@@ -0,0 +1,85 @@
+# YOLOv3
+
+## Model Zoo
+
+### YOLOv3 on COCO
+
+| 骨架网络             | 输入尺寸   | 每张GPU图片个数 | 学习率策略 |推理时间(fps) | mAP<sup>val<br>0.5:0.95 |                           下载                          | 配置文件 |
+| :------------------- | :------- | :-----: | :-----: | :------------: | :-----: | :-----------------------------------------------------: | :-----: |
+| DarkNet53(paper)  | 608         |    8    |   270e    |     -     |  33.0  |    -   |    -   |
+| DarkNet53(paper)  | 416         |    8    |   270e    |     -     |  31.0  |    -   |    -   |
+| DarkNet53(paper)  | 320         |    8    |   270e    |     -     |  28.2  |    -   |    -   |
+| DarkNet53         | 608         |    8    |   270e    |     -     |  **39.1**  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_coco.pdparams) | [配置文件](./yolov3_darknet53_270e_coco.yml) |
+| DarkNet53         | 416         |    8    |   270e    |     -     |  37.7  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_coco.pdparams) | [配置文件](./yolov3_darknet53_270e_coco.yml) |
+| DarkNet53         | 320         |    8    |   270e    |     -     |  34.8  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_coco.pdparams) | [配置文件](./yolov3_darknet53_270e_coco.yml) |
+|   ResNet50_vd-DCN    | 608        |    8    |   270e    |     -     |  **40.6**  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_r50vd_dcn_270e_coco.pdparams) | [配置文件](./yolov3_r50vd_dcn_270e_coco.yml) |
+|   ResNet50_vd-DCN    | 416        |    8    |   270e    |     -     |  38.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_r50vd_dcn_270e_coco.pdparams) | [配置文件](./yolov3_r50vd_dcn_270e_coco.yml) |
+|   ResNet50_vd-DCN    | 320        |    8    |   270e    |     -     |  35.1  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_r50vd_dcn_270e_coco.pdparams) | [配置文件](./yolov3_r50vd_dcn_270e_coco.yml) |
+| ResNet34         | 608         |    8    |   270e    |     -     |  36.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_r34_270e_coco.pdparams) | [配置文件](./yolov3_r34_270e_coco.yml) |
+| ResNet34         | 416         |    8    |   270e    |     -     |  34.3  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_r34_270e_coco.pdparams) | [配置文件](./yolov3_r34_270e_coco.yml) |
+| ResNet34         | 320         |    8    |   270e    |     -     |  31.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_r34_270e_coco.pdparams) | [配置文件](./yolov3_r34_270e_coco.yml) |
+| MobileNet-V1         | 608         |    8    |   270e    |     -     |  29.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v1_270e_coco.yml) |
+| MobileNet-V1         | 416         |    8    |   270e    |     -     |  29.3  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v1_270e_coco.yml) |
+| MobileNet-V1         | 320         |    8    |   270e    |     -     |  27.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v1_270e_coco.yml) |
+| MobileNet-V3         | 608         |    8    |   270e    |     -     |  31.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_270e_coco.yml) |
+| MobileNet-V3         | 416         |    8    |   270e    |     -     |  29.6  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_270e_coco.yml) |
+| MobileNet-V3         | 320         |    8    |   270e    |     -     |  27.1  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_270e_coco.yml) |
+| MobileNet-V1-SSLD    | 608         |    8    |   270e    |     -     |  31.0  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_ssld_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v1_ssld_270e_coco.yml) |
+| MobileNet-V1-SSLD    | 416         |    8    |   270e    |     -     |  30.6  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_ssld_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v1_ssld_270e_coco.yml) |
+| MobileNet-V1-SSLD    | 320         |    8    |   270e    |     -     |  28.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_ssld_270e_coco.pdparams) | [配置文件](./yolov3_mobilenet_v1_ssld_270e_coco.yml) |
+
+### YOLOv3 on Pasacl VOC
+
+| 骨架网络     | 输入尺寸 | 每张GPU图片个数 | 学习率策略 |推理时间(fps)| mAP(0.50,11point) | 下载 | 配置文件 |
+| :----------- | :--: | :-----: | :-----: |:------------: |:----: | :-------: | :----: |
+| DarkNet53    | 608  |    8    |   270e  |      -        |  **85.4** (56.1 mAP<br>0.5:0.95)  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_voc.pdparams) | [配置文件](./yolov3_darknet53_270e_voc.yml) |
+| DarkNet53    | 416  |    8    |   270e  |      -        |  85.2 (57.3 mAP<br>0.5:0.95)  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_voc.pdparams) | [配置文件](./yolov3_darknet53_270e_voc.yml) |
+| DarkNet53    | 320  |    8    |   270e  |      -        |  84.3 (55.2 mAP<br>0.5:0.95)  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_voc.pdparams) | [配置文件](./yolov3_darknet53_270e_voc.yml) |
+| MobileNet-V1 | 608  |    8    |   270e  |      -        |  75.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v1_270e_voc.yml) |
+| MobileNet-V1 | 416  |    8    |   270e  |      -        |  76.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v1_270e_voc.yml) |
+| MobileNet-V1 | 320  |    8    |   270e  |      -        |  74.3  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v1_270e_voc.yml) |
+| MobileNet-V3 | 608  |    8    |   270e  |      -        |  79.6  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_270e_voc.yml) |
+| MobileNet-V3 | 416  |    8    |   270e  |      -        |  78.6  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_270e_voc.yml) |
+| MobileNet-V3 | 320  |    8    |   270e  |      -        |  76.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_270e_voc.yml) |
+| MobileNet-V1-SSLD | 608  |    8    |   270e  |      -        |  78.3  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_ssld_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v1_ssld_270e_voc.yml) |
+| MobileNet-V1-SSLD | 416  |    8    |   270e  |      -        |  79.6  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_ssld_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v1_ssld_270e_voc.yml) |
+| MobileNet-V1-SSLD | 320  |    8    |   270e  |      -        |  77.3  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_ssld_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v1_ssld_270e_voc.yml) |
+| MobileNet-V3-SSLD | 608  |    8    |   270e  |      -        |  80.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_ssld_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_ssld_270e_voc.yml) |
+| MobileNet-V3-SSLD | 416  |    8    |   270e  |      -        |  79.2  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_ssld_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_ssld_270e_voc.yml) |
+| MobileNet-V3-SSLD | 320  |    8    |   270e  |      -        |  77.3  | [下载链接](https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v3_large_ssld_270e_voc.pdparams) | [配置文件](./yolov3_mobilenet_v3_large_ssld_270e_voc.yml) |
+
+
+**注意:**
+  - YOLOv3模型训练过程中默认使用8 GPUs，总batch_size默认为64，评估时网络尺度默认为`608*608`；
+  - `416*416`和`320*320`尺度只需更改`EvalReader`的`Resize`参数为相应值即可，无需重新训练模型，如：
+  ```
+    EvalReader:
+    sample_transforms:
+        - Decode: {}
+        - Resize: {target_size: [416, 416], keep_ratio: False, interp: 2} # or [320, 320]
+        - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+        - Permute: {}
+    batch_size: 1
+  ```
+  - VOC数据集可以从此[链接](https://bj.bcebos.com/v1/paddledet/data/voc.zip)下载，默认评估指标为mAP(0.50,11point)，如果想转为COCO格式指标的mAP<br>0.5:0.95，可以参照[yolov3_darknet53_270e_voc](./yolov3_darknet53_270e_voc.yml) 添加以下几行重新eval：
+  ```
+    metric: COCO
+    EvalDataset:
+    !COCODataSet
+        image_dir: VOCdevkit/VOC2007/JPEGImages
+        anno_path: voc_test.json
+        dataset_dir: dataset/voc
+  ```
+
+
+## Citations
+```
+@misc{redmon2018yolov3,
+    title={YOLOv3: An Incremental Improvement},
+    author={Joseph Redmon and Ali Farhadi},
+    year={2018},
+    eprint={1804.02767},
+    archivePrefix={arXiv},
+    primaryClass={cs.CV}
+}
+```
--- a/paddle_detection/configs/yolov3/_base_/optimizer_270e.yml
+++ b/paddle_detection/configs/yolov3/_base_/optimizer_270e.yml
@@ -0,0 +1,21 @@
+epoch: 270
+
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones:
+    - 216
+    - 243
+  - !LinearWarmup
+    start_factor: 0.
+    steps: 4000
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0005
+    type: L2
--- a/paddle_detection/configs/yolov3/_base_/optimizer_40e.yml
+++ b/paddle_detection/configs/yolov3/_base_/optimizer_40e.yml
@@ -0,0 +1,21 @@
+epoch: 40
+
+LearningRate:
+  base_lr: 0.0001
+  schedulers:
+  - name: PiecewiseDecay
+    gamma: 0.1
+    milestones:
+    - 32
+    - 36
+  - name: LinearWarmup
+    start_factor: 0.3333333333333333
+    steps: 100
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0005
+    type: L2
--- a/paddle_detection/configs/yolov3/_base_/yolov3_darknet53.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_darknet53.yml
@@ -0,0 +1,41 @@
+architecture: YOLOv3
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/DarkNet53_pretrained.pdparams
+norm_type: sync_bn
+
+YOLOv3:
+  backbone: DarkNet
+  neck: YOLOv3FPN
+  yolo_head: YOLOv3Head
+  post_process: BBoxPostProcess
+
+DarkNet:
+  depth: 53
+  return_idx: [2, 3, 4]
+
+# use default config
+# YOLOv3FPN:
+
+YOLOv3Head:
+  anchors: [[10, 13], [16, 30], [33, 23],
+            [30, 61], [62, 45], [59, 119],
+            [116, 90], [156, 198], [373, 326]]
+  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
+  loss: YOLOv3Loss
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+BBoxPostProcess:
+  decode:
+    name: YOLOBox
+    conf_thresh: 0.005
+    downsample_ratio: 32
+    clip_bbox: true
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 1000
--- a/paddle_detection/configs/yolov3/_base_/yolov3_mobilenet_v1.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_mobilenet_v1.yml
@@ -0,0 +1,43 @@
+architecture: YOLOv3
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/MobileNetV1_pretrained.pdparams
+norm_type: sync_bn
+
+YOLOv3:
+  backbone: MobileNet
+  neck: YOLOv3FPN
+  yolo_head: YOLOv3Head
+  post_process: BBoxPostProcess
+
+MobileNet:
+  scale: 1
+  feature_maps: [4, 6, 13]
+  with_extra_blocks: false
+  extra_block_filters: []
+
+# use default config
+# YOLOv3FPN:
+
+YOLOv3Head:
+  anchors: [[10, 13], [16, 30], [33, 23],
+            [30, 61], [62, 45], [59, 119],
+            [116, 90], [156, 198], [373, 326]]
+  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
+  loss: YOLOv3Loss
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+BBoxPostProcess:
+  decode:
+    name: YOLOBox
+    conf_thresh: 0.005
+    downsample_ratio: 32
+    clip_bbox: true
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 1000
--- a/paddle_detection/configs/yolov3/_base_/yolov3_mobilenet_v3_large.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_mobilenet_v3_large.yml
@@ -0,0 +1,44 @@
+architecture: YOLOv3
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/MobileNetV3_large_x1_0_ssld_pretrained.pdparams
+norm_type: sync_bn
+
+YOLOv3:
+  backbone: MobileNetV3
+  neck: YOLOv3FPN
+  yolo_head: YOLOv3Head
+  post_process: BBoxPostProcess
+
+MobileNetV3:
+  model_name: large
+  scale: 1.
+  with_extra_blocks: false
+  extra_block_filters: []
+  feature_maps: [7, 13, 16]
+
+# use default config
+# YOLOv3FPN:
+
+YOLOv3Head:
+  anchors: [[10, 13], [16, 30], [33, 23],
+            [30, 61], [62, 45], [59, 119],
+            [116, 90], [156, 198], [373, 326]]
+  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
+  loss: YOLOv3Loss
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+BBoxPostProcess:
+  decode:
+    name: YOLOBox
+    conf_thresh: 0.005
+    downsample_ratio: 32
+    clip_bbox: true
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 1000
--- a/paddle_detection/configs/yolov3/_base_/yolov3_mobilenet_v3_small.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_mobilenet_v3_small.yml
@@ -0,0 +1,44 @@
+architecture: YOLOv3
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/MobileNetV3_small_x1_0_ssld_pretrained.pdparams
+norm_type: sync_bn
+
+YOLOv3:
+  backbone: MobileNetV3
+  neck: YOLOv3FPN
+  yolo_head: YOLOv3Head
+  post_process: BBoxPostProcess
+
+MobileNetV3:
+  model_name: small
+  scale: 1.
+  with_extra_blocks: false
+  extra_block_filters: []
+  feature_maps: [4, 9, 12]
+
+# use default config
+# YOLOv3FPN:
+
+YOLOv3Head:
+  anchors: [[10, 13], [16, 30], [33, 23],
+            [30, 61], [62, 45], [59, 119],
+            [116, 90], [156, 198], [373, 326]]
+  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
+  loss: YOLOv3Loss
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+BBoxPostProcess:
+  decode:
+    name: YOLOBox
+    conf_thresh: 0.005
+    downsample_ratio: 32
+    clip_bbox: true
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 1000
--- a/paddle_detection/configs/yolov3/_base_/yolov3_r34.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_r34.yml
@@ -0,0 +1,41 @@
+architecture: YOLOv3
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/ResNet34_pretrained.pdparams
+norm_type: sync_bn
+
+YOLOv3:
+  backbone: ResNet
+  neck: YOLOv3FPN
+  yolo_head: YOLOv3Head
+  post_process: BBoxPostProcess
+
+ResNet:
+  depth: 34
+  return_idx: [1, 2, 3]
+  freeze_at: -1
+  freeze_norm: false
+  norm_decay: 0.
+
+YOLOv3Head:
+  anchors: [[10, 13], [16, 30], [33, 23],
+            [30, 61], [62, 45], [59, 119],
+            [116, 90], [156, 198], [373, 326]]
+  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
+  loss: YOLOv3Loss
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+BBoxPostProcess:
+  decode:
+    name: YOLOBox
+    conf_thresh: 0.005
+    downsample_ratio: 32
+    clip_bbox: true
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 1000
--- a/paddle_detection/configs/yolov3/_base_/yolov3_r50vd_dcn.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_r50vd_dcn.yml
@@ -0,0 +1,45 @@
+architecture: YOLOv3
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/ResNet50_vd_ssld_pretrained.pdparams
+norm_type: sync_bn
+
+YOLOv3:
+  backbone: ResNet
+  neck: YOLOv3FPN
+  yolo_head: YOLOv3Head
+  post_process: BBoxPostProcess
+
+ResNet:
+  depth: 50
+  variant: d
+  return_idx: [1, 2, 3]
+  dcn_v2_stages: [3]
+  freeze_at: -1
+  freeze_norm: false
+  norm_decay: 0.
+
+# YOLOv3FPN:
+
+YOLOv3Head:
+  anchors: [[10, 13], [16, 30], [33, 23],
+            [30, 61], [62, 45], [59, 119],
+            [116, 90], [156, 198], [373, 326]]
+  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
+  loss: YOLOv3Loss
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+BBoxPostProcess:
+  decode:
+    name: YOLOBox
+    conf_thresh: 0.005
+    downsample_ratio: 32
+    clip_bbox: true
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 1000
--- a/paddle_detection/configs/yolov3/_base_/yolov3_reader.yml
+++ b/paddle_detection/configs/yolov3/_base_/yolov3_reader.yml
@@ -0,0 +1,44 @@
+worker_num: 2
+TrainReader:
+  inputs_def:
+    num_max_boxes: 50
+  sample_transforms:
+    - Decode: {}
+    - Mixup: {alpha: 1.5, beta: 1.5}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53]}
+    - RandomCrop: {}
+    - RandomFlip: {}
+  batch_transforms:
+    - BatchRandomResize: {target_size: [320, 352, 384, 416, 448, 480, 512, 544, 576, 608], random_size: True, random_interp: True, keep_ratio: False}
+    - NormalizeBox: {}
+    - PadBox: {num_max_boxes: 50}
+    - BboxXYXY2XYWH: {}
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+    - Gt2YoloTarget: {anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]], anchors: [[10, 13], [16, 30], [33, 23], [30, 61], [62, 45], [59, 119], [116, 90], [156, 198], [373, 326]], downsample_ratios: [32, 16, 8]}
+  batch_size: 8
+  shuffle: true
+  drop_last: true
+  mixup_epoch: 250
+  use_shared_memory: true
+
+EvalReader:
+  inputs_def:
+    num_max_boxes: 50
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [608, 608], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+  batch_size: 1
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 608, 608]
+  sample_transforms:
+    - Decode: {}
+    - Resize: {target_size: [608, 608], keep_ratio: False, interp: 2}
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+  batch_size: 1
--- a/paddle_detection/configs/yolov3/yolov3_darknet53_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_darknet53_270e_coco.yml
@@ -0,0 +1,10 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_darknet53.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_darknet53_270e_coco/model_final
--- a/paddle_detection/configs/yolov3/yolov3_darknet53_270e_voc.yml
+++ b/paddle_detection/configs/yolov3/yolov3_darknet53_270e_voc.yml
@@ -0,0 +1,28 @@
+_BASE_: [
+  '../datasets/voc.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_darknet53.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_darknet53_270e_voc/model_final
+pretrain_weights: https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_coco.pdparams
+
+
+# set collate_batch to false because ground-truth info is needed
+# on voc dataset and should not collate data in batch when batch size
+# is larger than 1.
+EvalReader:
+  collate_batch: false
+
+
+# ### remove comment below and run evaluate again to get 56.1 COCO for mAP(0.5:0.95)
+# metric: COCO
+# EvalDataset:
+#   !COCODataSet
+#     image_dir: VOCdevkit/VOC2007/JPEGImages
+#     anno_path: voc_test.json
+#     dataset_dir: dataset/voc
+#     # wget https://bj.bcebos.com/v1/paddledet/data/voc.zip
--- a/paddle_detection/configs/yolov3/yolov3_darknet53_original_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_darknet53_original_270e_coco.yml
@@ -0,0 +1,40 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_darknet53.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_darknet53_270e_coco/model_final
+
+norm_type: bn
+
+YOLOv3Loss:
+  ignore_thresh: 0.5
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+TrainReader:
+  inputs_def:
+    num_max_boxes: 50
+  sample_transforms:
+    - Decode: {}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53], ratio: 2.0}
+    - RandomCrop: {}
+    - RandomFlip: {}
+  batch_transforms:
+    - BatchRandomResize: {target_size: [320, 352, 384, 416, 448, 480, 512, 544, 576, 608], random_size: True, random_interp: True, keep_ratio: False}
+    - NormalizeBox: {}
+    - PadBox: {num_max_boxes: 50}
+    - BboxXYXY2XYWH: {}
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+    - Gt2YoloTarget: {anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]], anchors: [[10, 13], [16, 30], [33, 23], [30, 61], [62, 45], [59, 119], [116, 90], [156, 198], [373, 326]], downsample_ratios: [32, 16, 8], iou_thresh: 0.5}
+  batch_size: 8
+  shuffle: true
+  drop_last: true
+  mixup_epoch: -1
+  use_shared_memory: true
--- a/paddle_detection/configs/yolov3/yolov3_darknet53_original_320e_coco_1p.yml
+++ b/paddle_detection/configs/yolov3/yolov3_darknet53_original_320e_coco_1p.yml
@@ -0,0 +1,59 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/yolov3_darknet53.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_darknet53_270e_coco/model_final
+
+norm_type: bn
+
+YOLOv3Loss:
+  ignore_thresh: 0.5
+  downsample: [32, 16, 8]
+  label_smooth: false
+
+worker_num: 8
+TrainReader:
+  inputs_def:
+    num_max_boxes: 50
+  sample_transforms:
+    - Decode: {}
+    - RandomDistort: {}
+    - RandomExpand: {fill_value: [123.675, 116.28, 103.53], ratio: 2.0}
+    - RandomCrop: {}
+    - RandomFlip: {}
+  batch_transforms:
+    - BatchRandomResize: {target_size: [416], random_size: True, random_interp: True, keep_ratio: False}
+    - NormalizeBox: {}
+    - PadBox: {num_max_boxes: 50}
+    - BboxXYXY2XYWH: {}
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+    - Gt2YoloTarget: {anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]], anchors: [[10, 13], [16, 30], [33, 23], [30, 61], [62, 45], [59, 119], [116, 90], [156, 198], [373, 326]], downsample_ratios: [32, 16, 8], iou_thresh: 0.5}
+  batch_size: 32
+  shuffle: true
+  drop_last: true
+  mixup_epoch: -1
+  use_shared_memory: true
+
+epoch: 320
+
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+  - !CosineDecay
+    max_epochs: 320
+  - !LinearWarmup
+    start_factor: 0.
+    epochs: 4
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.016
+    type: L2
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_270e_coco.yml
@@ -0,0 +1,10 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v1.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_mobilenet_v1_270e_coco/model_final
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_270e_voc.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_270e_voc.yml
@@ -0,0 +1,28 @@
+_BASE_: [
+  '../datasets/voc.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v1.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_mobilenet_v1_270e_voc/model_final
+
+# set collate_batch to false because ground-truth info is needed
+# on voc dataset and should not collate data in batch when batch size
+# is larger than 1.
+EvalReader:
+  collate_batch: false
+
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones:
+    - 216
+    - 243
+  - !LinearWarmup
+    start_factor: 0.
+    steps: 1000
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_roadsign.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_roadsign.yml
@@ -0,0 +1,13 @@
+_BASE_: [
+  '../datasets/roadsign_voc.yml',
+  '../runtime.yml',
+  '_base_/optimizer_40e.yml',
+  '_base_/yolov3_mobilenet_v1.yml',
+  '_base_/yolov3_reader.yml',
+]
+pretrain_weights: https://paddledet.bj.bcebos.com/models/yolov3_mobilenet_v1_270e_coco.pdparams
+weights: output/yolov3_mobilenet_v1_roadsign/model_final
+
+YOLOv3Loss:
+  ignore_thresh: 0.7
+  label_smooth: true
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_ssld_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_ssld_270e_coco.yml
@@ -0,0 +1,11 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v1.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+pretrain_weights:  https://paddledet.bj.bcebos.com/models/pretrained/MobileNetV1_ssld_pretrained.pdparams
+weights: output/yolov3_mobilenet_v1_ssld_270e_coco/model_final
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_ssld_270e_voc.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v1_ssld_270e_voc.yml
@@ -0,0 +1,29 @@
+_BASE_: [
+  '../datasets/voc.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v1.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+pretrain_weights:  https://paddledet.bj.bcebos.com/models/pretrained/MobileNetV1_ssld_pretrained.pdparams
+weights: output/yolov3_mobilenet_v1_ssld_270e_voc/model_final
+
+# set collate_batch to false because ground-truth info is needed
+# on voc dataset and should not collate data in batch when batch size
+# is larger than 1.
+EvalReader:
+  collate_batch: false
+
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones:
+    - 216
+    - 243
+  - !LinearWarmup
+    start_factor: 0.
+    steps: 1000
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v3_large_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v3_large_270e_coco.yml
@@ -0,0 +1,10 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v3_large.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_mobilenet_v3_large_270e_coco/model_final
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v3_large_270e_voc.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v3_large_270e_voc.yml
@@ -0,0 +1,28 @@
+_BASE_: [
+  '../datasets/voc.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v3_large.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_mobilenet_v3_large_270e_voc/model_final
+
+# set collate_batch to false because ground-truth info is needed
+# on voc dataset and should not collate data in batch when batch size
+# is larger than 1.
+EvalReader:
+  collate_batch: false
+
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones:
+    - 216
+    - 243
+  - !LinearWarmup
+    start_factor: 0.
+    steps: 1000
--- a/paddle_detection/configs/yolov3/yolov3_mobilenet_v3_large_ssld_270e_voc.yml
+++ b/paddle_detection/configs/yolov3/yolov3_mobilenet_v3_large_ssld_270e_voc.yml
@@ -0,0 +1,29 @@
+_BASE_: [
+  '../datasets/voc.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_mobilenet_v3_large.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+pretrain_weights:  https://paddledet.bj.bcebos.com/models/pretrained/MobileNetV3_large_x1_0_ssld_pretrained.pdparams
+weights: output/yolov3_mobilenet_v3_large_ssld_270e_voc/model_final
+
+# set collate_batch to false because ground-truth info is needed
+# on voc dataset and should not collate data in batch when batch size
+# is larger than 1.
+EvalReader:
+  collate_batch: false
+
+LearningRate:
+  base_lr: 0.001
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones:
+    - 216
+    - 243
+  - !LinearWarmup
+    start_factor: 0.
+    steps: 1000
--- a/paddle_detection/configs/yolov3/yolov3_r34_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_r34_270e_coco.yml
@@ -0,0 +1,10 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_r34.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_r34_270e_coco/model_final
--- a/paddle_detection/configs/yolov3/yolov3_r50vd_dcn_270e_coco.yml
+++ b/paddle_detection/configs/yolov3/yolov3_r50vd_dcn_270e_coco.yml
@@ -0,0 +1,10 @@
+_BASE_: [
+  '../datasets/coco_detection.yml',
+  '../runtime.yml',
+  '_base_/optimizer_270e.yml',
+  '_base_/yolov3_r50vd_dcn.yml',
+  '_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: output/yolov3_r50vd_dcn_270e_coco/model_final