Bump version to v2.22.0

Bump version to v2.22.0

Author: ZwwWayne

.dev_scripts/benchmark_inference_fps.py

@@ -7,6 +7,7 @@
 from mmcv import Config, DictAction
 from mmcv.runner import init_dist
 from terminaltables import GithubFlavoredMarkdownTable
+
 from tools.analysis_tools.benchmark import repeat_measure_inference_speed
 
 

.dev_scripts/gather_models.py

@@ -98,7 +98,9 @@ def get_dataset_name(config):
         LVISV05Dataset='LVIS v0.5',
         LVISV1Dataset='LVIS v1',
         VOCDataset='Pascal VOC',
-        WIDERFaceDataset='WIDER Face')
+        WIDERFaceDataset='WIDER Face',
+        OpenImagesDataset='OpenImagesDataset',
+        OpenImagesChallengeDataset='OpenImagesChallengeDataset')
     cfg = mmcv.Config.fromfile('./configs/' + config)
     return name_map[cfg.dataset_type]
 

.github/workflows/build.yml

@@ -33,23 +33,26 @@ jobs:
     strategy:
       matrix:
         python-version: [3.7]
-        torch: [1.5.1, 1.6.0, 1.7.0, 1.8.0, 1.9.0]
+        torch: [1.5.1, 1.6.0, 1.7.0, 1.8.0, 1.9.0, 1.10.1]
         include:
           - torch: 1.5.1
             torchvision: 0.6.1
-            mmcv: 1.5.0
+            mmcv: 1.5
           - torch: 1.6.0
             torchvision: 0.7.0
-            mmcv: 1.6.0
+            mmcv: 1.6
           - torch: 1.7.0
             torchvision: 0.8.1
-            mmcv: 1.7.0
+            mmcv: 1.7
           - torch: 1.8.0
             torchvision: 0.9.0
-            mmcv: 1.8.0
+            mmcv: 1.8
           - torch: 1.9.0
             torchvision: 0.10.0
-            mmcv: 1.9.0
+            mmcv: 1.9
+          - torch: 1.10.1
+            torchvision: 0.11.2
+            mmcv: 1.10
     steps:
       - uses: actions/checkout@v2
       - name: Set up Python ${{ matrix.python-version }}
@@ -91,19 +94,19 @@ jobs:
           - torch: 1.5.1+cu101
             torch_version: torch1.5.1
             torchvision: 0.6.1+cu101
-            mmcv: 1.5.0
+            mmcv: 1.5
           - torch: 1.6.0+cu101
             torch_version: torch1.6.0
             torchvision: 0.7.0+cu101
-            mmcv: 1.6.0
+            mmcv: 1.6
           - torch: 1.7.0+cu101
             torch_version: torch1.7.0
             torchvision: 0.8.1+cu101
-            mmcv: 1.7.0
+            mmcv: 1.7
           - torch: 1.8.0+cu101
             torch_version: torch1.8.0
             torchvision: 0.9.0+cu101
-            mmcv: 1.8.0
+            mmcv: 1.8
 
     steps:
       - uses: actions/checkout@v2
@@ -160,12 +163,16 @@ jobs:
     strategy:
       matrix:
         python-version: [3.6, 3.7, 3.8, 3.9]
-        torch: [1.9.0+cu102]
+        torch: [1.9.0+cu102, 1.10.1+cu102]
         include:
           - torch: 1.9.0+cu102
             torch_version: torch1.9.0
             torchvision: 0.10.0+cu102
-            mmcv: 1.9.0
+            mmcv: 1.9
+          - torch: 1.10.1+cu102
+            torch_version: torch1.10.1
+            torchvision: 0.11.2+cu102
+            mmcv: 1.10
 
     steps:
       - uses: actions/checkout@v2
@@ -224,3 +231,50 @@ jobs:
           env_vars: OS,PYTHON
           name: codecov-umbrella
           fail_ci_if_error: false
+
+  build_windows:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [windows-2022]
+        python: [3.8]
+        platform: [cpu, cu111]
+    steps:
+      - uses: actions/checkout@v2
+      - name: Set up Python ${{ matrix.python }}
+        uses: actions/setup-python@v2
+        with:
+          python-version: ${{ matrix.python }}
+      - name: Upgrade pip
+        run: pip install pip --upgrade --user
+      - name: Install PyTorch
+        # As a complement to Linux CI, we test on PyTorch LTS version
+        run: pip install torch==1.8.2+${{ matrix.platform }} torchvision==0.9.2+${{ matrix.platform }} -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html
+      - name: Install MMCV
+        run: pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.8/index.html --only-binary mmcv-full
+      - name: Install unittest dependencies
+        run: |
+          python -V
+          python -m pip install pycocotools
+          python -m pip install -r requirements/tests.txt -r requirements/optional.txt
+          python -m pip install albumentations>=0.3.2 --no-binary imgaug,albumentations
+          python -m pip install git+https://github.com/cocodataset/panopticapi.git
+          python -c 'import mmcv; print(mmcv.__version__)'
+      - name: Show pip list
+        run: pip list
+      - name: Build and install
+        run: pip install -e .
+      - name: Run unittests
+        run: coverage run --branch --source mmdet -m pytest tests -sv
+      - name: Generate coverage report
+        run: |
+          coverage xml
+          coverage report -m
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v2
+        with:
+          file: ./coverage.xml
+          flags: unittests
+          env_vars: OS,PYTHON
+          name: codecov-umbrella
+          fail_ci_if_error: false

.owners.yml

@@ -0,0 +1,14 @@
+assign:
+  strategy:
+    # random
+    daily-shift-based
+  scedule:
+    '*/1 * * * *'
+  assignees:
+    - Czm369
+    - hhaAndroid
+    - jbwang1997
+    - RangiLyu
+    - BIGWangYuDong
+    - chhluo
+    - ZwwWayne

.pre-commit-config.yaml

@@ -3,12 +3,8 @@ repos:
     rev: 3.8.3
     hooks:
       - id: flake8
-  - repo: https://github.com/asottile/seed-isort-config
-    rev: v2.2.0
-    hooks:
-      - id: seed-isort-config
-  - repo: https://github.com/timothycrosley/isort
-    rev: 4.3.21
+  - repo: https://github.com/PyCQA/isort
+    rev: 5.10.1
     hooks:
       - id: isort
   - repo: https://github.com/pre-commit/mirrors-yapf

README.md

@@ -74,10 +74,11 @@ This project is released under the [Apache 2.0 license](LICENSE).
 
 ## Changelog
 
-**2.21.0** was released in 8/2/2022:
+**2.22.0** was released in 24/2/2022:
 
-- Support CPU training
-- Allow to set parameters about multi-processing to speed up training and testing
+- Support [MaskFormer](configs/maskformer), [DyHead](configs/dyhead), [OpenImages Dataset](configs/openimages) and [TIMM backbone](configs/timm_example)
+- Support visualization for Panoptic Segmentation
+- Release a good recipe of using ResNet in object detectors pre-trained by [ResNet Strikes Back](https://arxiv.org/abs/2110.00476), which consistently brings about 3~4 mAP improvements over RetinaNet, Faster/Mask/Cascade Mask R-CNN
 
 Please refer to [changelog.md](docs/en/changelog.md) for details and release history.
 
@@ -162,6 +163,7 @@ Results and models are available in the [model zoo](docs/en/model_zoo.md).
       <td>
         <ul>
           <li><a href="configs/panoptic_fpn">Panoptic FPN (CVPR'2019)</a></li>
+          <li><a href="configs/maskformer">MaskFormer (NeurIPS'2019)</a></li>
         </ul>
       </td>
       <td>
@@ -225,6 +227,7 @@ Results and models are available in the [model zoo](docs/en/model_zoo.md).
         <li><a href="configs/pvt">PVT (ICCV'2021)</a></li>
         <li><a href="configs/swin">Swin (CVPR'2021)</a></li>
         <li><a href="configs/pvt">PVTv2 (ArXiv'2021)</a></li>
+        <li><a href="configs/resnet_strikes_back">ResNet strikes back (ArXiv'2021)</a></li>
       </ul>
       </td>
       <td>
@@ -234,6 +237,7 @@ Results and models are available in the [model zoo](docs/en/model_zoo.md).
         <li><a href="configs/carafe">CARAFE (ICCV'2019)</a></li>
         <li><a href="configs/fpg">FPG (ArXiv'2020)</a></li>
         <li><a href="configs/groie">GRoIE (ICPR'2020)</a></li>
+        <li><a href="configs/dyhead">DyHead (CVPR'2021)</a></li>
       </ul>
       </td>
       <td>
@@ -252,6 +256,7 @@ Results and models are available in the [model zoo](docs/en/model_zoo.md).
           <li><a href="configs/gn+ws">Weight Standardization (ArXiv'2019)</a></li>
           <li><a href="configs/pisa">Prime Sample Attention (CVPR'2020)</a></li>
           <li><a href="configs/strong_baselines">Strong Baselines (CVPR'2021)</a></li>
+          <li><a href="configs/resnet_strikes_back">Resnet strikes back (ArXiv'2021)</a></li>
         </ul>
       </td>
     </tr>
@@ -321,4 +326,4 @@ If you use this toolbox or benchmark in your research, please cite this project.
 - [MMFlow](https://github.com/open-mmlab/mmflow): OpenMMLab optical flow toolbox and benchmark.
 - [MMEditing](https://github.com/open-mmlab/mmediting): OpenMMLab image and video editing toolbox.
 - [MMGeneration](https://github.com/open-mmlab/mmgeneration): OpenMMLab image and video generative models toolbox.
-- [MMDeploy](https://github.com/open-mmlab/mmdeploy): OpenMMLab Model Deployment Framework.
+- [MMDeploy](https://github.com/open-mmlab/mmdeploy): OpenMMLab model deployment framework.

README_zh-CN.md

@@ -73,10 +73,11 @@ MMDetection 是一个基于 PyTorch 的目标检测开源工具箱。它是 [Ope
 
 ## 更新日志
 
-最新的 **2.21.0** 版本已经在 2022.02.08 发布:
+最新的 **2.22.0** 版本已经在 2022.02.24 发布:
 
-- 支持了 CPU 训练
-- 允许设置多进程相关的参数来加速训练与推理
+- 支持 [MaskFormer](configs/maskformer)，[DyHead](configs/dyhead)，[OpenImages Dataset](configs/openimages) 和 [TIMM backbone](configs/timm_example)
+- 支持全景分割可视化
+- 发布了一个在目标检测任务中使用 ResNet 的好方法，它是由 [ResNet Strikes Back](https://arxiv.org/abs/2110.00476) 预训练的，并且能稳定的在 RetinaNet, Faster/Mask/Cascade Mask R-CNN 上带来约 3-4 mAP 的提升
 
 如果想了解更多版本更新细节和历史信息，请阅读[更新日志](docs/changelog.md)。
 
@@ -224,6 +225,7 @@ MMDetection 是一个基于 PyTorch 的目标检测开源工具箱。它是 [Ope
         <li><a href="configs/pvt">PVT (ICCV'2021)</a></li>
         <li><a href="configs/swin">Swin (CVPR'2021)</a></li>
         <li><a href="configs/pvt">PVTv2 (ArXiv'2021)</a></li>
+        <li><a href="configs/resnet_strikes_back">ResNet strikes back (ArXiv'2021)</a></li>
       </ul>
       </td>
       <td>
@@ -233,6 +235,7 @@ MMDetection 是一个基于 PyTorch 的目标检测开源工具箱。它是 [Ope
         <li><a href="configs/carafe">CARAFE (ICCV'2019)</a></li>
         <li><a href="configs/fpg">FPG (ArXiv'2020)</a></li>
         <li><a href="configs/groie">GRoIE (ICPR'2020)</a></li>
+        <li><a href="configs/dyhead">DyHead (CVPR'2021)</a></li>
       </ul>
       </td>
       <td>
@@ -251,6 +254,7 @@ MMDetection 是一个基于 PyTorch 的目标检测开源工具箱。它是 [Ope
           <li><a href="configs/gn+ws">Weight Standardization (ArXiv'2019)</a></li>
           <li><a href="configs/pisa">Prime Sample Attention (CVPR'2020)</a></li>
           <li><a href="configs/strong_baselines">Strong Baselines (CVPR'2021)</a></li>
+          <li><a href="configs/resnet_strikes_back">Resnet strikes back (ArXiv'2021)</a></li>
         </ul>
       </td>
     </tr>

configs/_base_/datasets/openimages_detection.py

@@ -0,0 +1,65 @@
+# dataset settings
+dataset_type = 'OpenImagesDataset'
+data_root = 'data/OpenImages/'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='LoadAnnotations', with_bbox=True, denorm_bbox=True),
+    dict(type='Resize', img_scale=(1024, 800), keep_ratio=True),
+    dict(type='RandomFlip', flip_ratio=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='Pad', size_divisor=32),
+    dict(type='DefaultFormatBundle'),
+    dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']),
+]
+test_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(
+        type='MultiScaleFlipAug',
+        img_scale=(1024, 800),
+        flip=False,
+        transforms=[
+            dict(type='Resize', keep_ratio=True),
+            dict(type='RandomFlip'),
+            dict(type='Normalize', **img_norm_cfg),
+            dict(type='Pad', size_divisor=32),
+            dict(type='ImageToTensor', keys=['img']),
+            dict(type='Collect', keys=['img']),
+        ],
+    ),
+]
+data = dict(
+    samples_per_gpu=2,
+    workers_per_gpu=0,  # workers_per_gpu > 0 may occur out of memory
+    train=dict(
+        type=dataset_type,
+        ann_file=data_root + 'annotations/oidv6-train-annotations-bbox.csv',
+        img_prefix=data_root + 'OpenImages/train/',
+        label_file=data_root + 'annotations/class-descriptions-boxable.csv',
+        hierarchy_file=data_root +
+        'annotations/bbox_labels_600_hierarchy.json',
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        ann_file=data_root + 'annotations/validation-annotations-bbox.csv',
+        img_prefix=data_root + 'OpenImages/validation/',
+        label_file=data_root + 'annotations/class-descriptions-boxable.csv',
+        hierarchy_file=data_root +
+        'annotations/bbox_labels_600_hierarchy.json',
+        meta_file=data_root + 'annotations/validation-image-metas.pkl',
+        image_level_ann_file=data_root +
+        'annotations/validation-annotations-human-imagelabels-boxable.csv',
+        pipeline=test_pipeline),
+    test=dict(
+        type=dataset_type,
+        ann_file=data_root + 'annotations/validation-annotations-bbox.csv',
+        img_prefix=data_root + 'OpenImages/validation/',
+        label_file=data_root + 'annotations/class-descriptions-boxable.csv',
+        hierarchy_file=data_root +
+        'annotations/bbox_labels_600_hierarchy.json',
+        meta_file=data_root + 'annotations/validation-image-metas.pkl',
+        image_level_ann_file=data_root +
+        'annotations/validation-annotations-human-imagelabels-boxable.csv',
+        pipeline=test_pipeline))
+evaluation = dict(interval=1, metric='mAP')

configs/dyhead/README.md

@@ -0,0 +1,46 @@
+# DyHead
+
+> [Dynamic Head: Unifying Object Detection Heads with Attentions](https://arxiv.org/abs/2106.08322)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+The complex nature of combining localization and classification in object detection has resulted in the flourished development of methods. Previous works tried to improve the performance in various object detection heads but failed to present a unified view. In this paper, we present a novel dynamic head framework to unify object detection heads with attentions. By coherently combining multiple self-attention mechanisms between feature levels for scale-awareness, among spatial locations for spatial-awareness, and within output channels for task-awareness, the proposed approach significantly improves the representation ability of object detection heads without any computational overhead. Further experiments demonstrate that the effectiveness and efficiency of the proposed dynamic head on the COCO benchmark. With a standard ResNeXt-101-DCN backbone, we largely improve the performance over popular object detectors and achieve a new state-of-the-art at 54.0 AP. Furthermore, with latest transformer backbone and extra data, we can push current best COCO result to a new record at 60.6 AP.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/42844407/149169448-fcafb6d0-b866-41cc-9422-94de9f1e1761.png" height="300"/>
+</div>
+
+## Results and Models
+
+| Method | Backbone | Style   | Setting      | Lr schd | Mem (GB) | Inf time (fps) | box AP | Config | Download |
+|:------:|:--------:|:-------:|:------------:|:-------:|:--------:|:--------------:|:------:|:------:|:--------:|
+| ATSS   | R-50     | caffe   | reproduction | 1x      | 5.4      | 13.2           | 42.5   | [config](./atss_r50_caffe_fpn_dyhead_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/dyhead/atss_r50_fpn_dyhead_for_reproduction_1x_coco/atss_r50_fpn_dyhead_for_reproduction_4x4_1x_coco_20220107_213939-162888e6.pth) &#124; [log](https://download.openmmlab.com/mmdetection/v2.0/dyhead/atss_r50_fpn_dyhead_for_reproduction_1x_coco/atss_r50_fpn_dyhead_for_reproduction_4x4_1x_coco_20220107_213939.log.json) |
+| ATSS   | R-50     | pytorch | simple       | 1x      | 4.9      | 13.7           | 43.3   | [config](./atss_r50_fpn_dyhead_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/dyhead/atss_r50_fpn_dyhead_4x4_1x_coco/atss_r50_fpn_dyhead_4x4_1x_coco_20211219_023314-eaa620c6.pth) &#124; [log](https://download.openmmlab.com/mmdetection/v2.0/dyhead/atss_r50_fpn_dyhead_4x4_1x_coco/atss_r50_fpn_dyhead_4x4_1x_coco_20211219_023314.log.json) |
+
+- We trained the above models with 4 GPUs and 4 `samples_per_gpu`.
+- The `reproduction` setting aims to reproduce the official implementation based on Detectron2.
+- The `simple` setting serves as a minimum example to use DyHead in MMDetection. Specifically,
+  - it adds `DyHead` to `neck` after `FPN`
+  - it sets `stacked_convs=0` to `bbox_head`
+- The `simple` setting achieves higher AP than the original implementation.
+  We have not conduct ablation study between the two settings.
+  `dict(type='Pad', size_divisor=128)` may further improve AP by prefer spatial alignment across pyramid levels, although large padding reduces efficiency.
+
+## Relation to Other Methods
+
+- DyHead can be regarded as an improved [SEPC](https://arxiv.org/abs/2005.03101) with [DyReLU modules](https://arxiv.org/abs/2003.10027) and simplified [SE blocks](https://arxiv.org/abs/1709.01507).
+- Xiyang Dai et al., the author team of DyHead, adopt it for [Dynamic DETR](https://openaccess.thecvf.com/content/ICCV2021/html/Dai_Dynamic_DETR_End-to-End_Object_Detection_With_Dynamic_Attention_ICCV_2021_paper.html).
+  The description of Dynamic Encoder in Sec. 3.2 will help you understand DyHead.
+
+## Citation
+
+```latex
+@inproceedings{DyHead_CVPR2021,
+  author    = {Dai, Xiyang and Chen, Yinpeng and Xiao, Bin and Chen, Dongdong and Liu, Mengchen and Yuan, Lu and Zhang, Lei},
+  title     = {Dynamic Head: Unifying Object Detection Heads With Attentions},
+  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
+  year      = {2021}
+}
+```