open-mmlab · hhaAndroid · Oct 12, 2023 · Oct 12, 2023 · Oct 12, 2023 · Oct 12, 2023
diff --git a/README.md b/README.md
@@ -103,6 +103,52 @@ Apart from MMDetection, we also released [MMEngine](https://github.com/open-mmla
 
 ### Highlight
 
+**v3.2.0** was released in 12/10/2023:
+
+**1. Detection Transformer SOTA Model Collection**
+(1) Supported four updated and stronger SOTA Transformer models: [DDQ](configs/ddq/README.md), [CO-DETR](projects/CO-DETR/README.md), [AlignDETR](projects/AlignDETR/README.md), and [H-DINO](projects/HDINO/README.md).
+(2) Based on CO-DETR, MMDet released a model with a COCO performance of 64.1 mAP.
+(3) Algorithms such as DINO support `AMP/Checkpoint/FrozenBN`, which can effectively reduce memory usage.
+
+**2. [Comprehensive Performance Comparison between CNN and Transformer](<(projects/RF100-Benchmark/README.md)>)**
+RF100 consists of a dataset collection of 100 real-world datasets, including 7 domains. It can be used to assess the performance differences of Transformer models like DINO and CNN-based algorithms under different scenarios and data volumes. Users can utilize this benchmark to quickly evaluate the robustness of their algorithms in various scenarios.
+
+<div align=center>
+<img src="https://github.com/open-mmlab/mmdetection/assets/17425982/86420903-36a8-410d-9251-4304b9704f7d"/>
+</div>
+
+**3. Support for [GLIP](configs/glip/README.md) and [Grounding DINO](configs/grounding_dino/README.md) fine-tuning, the only algorithm library that supports Grounding DINO fine-tuning**
+The Grounding DINO algorithm in MMDet is the only library that supports fine-tuning. Its performance is one point higher than the official version, and of course, GLIP also outperforms the official version.
+We also provide a detailed process for training and evaluating Grounding DINO on custom datasets. Everyone is welcome to give it a try.
+
+|       Model        | Backbone |   Style   |  COCO mAP  | Official COCO mAP |
+| :----------------: | :------: | :-------: | :--------: | :---------------: |
+|  Grounding DINO-T  |  Swin-T  | Zero-shot |    48.5    |       48.4        |
+|  Grounding DINO-T  |  Swin-T  | Finetune  | 58.1(+0.9) |       57.2        |
+|  Grounding DINO-B  |  Swin-B  | Zero-shot |    56.9    |       56.7        |
+|  Grounding DINO-B  |  Swin-B  | Finetune  |    59.7    |                   |
+| Grounding DINO-R50 |   R50    |  Scratch  | 48.9(+0.8) |       48.1        |
+
+**4. Support for the open-vocabulary detection algorithm [Detic](projects/Detic_new/README.md) and multi-dataset joint training.**
+**5. Training detection models using [FSDP and DeepSpeed](<(projects/example_largemodel/README.md)>).**
+
+| ID  | AMP | GC of Backbone | GC of Encoder | FSDP | Peak Mem (GB) | Iter Time (s) |
+| :-: | :-: | :------------: | :-----------: | :--: | :-----------: | :-----------: |
+|  1  |     |                |               |      |   49 (A100)   |      0.9      |
+|  2  |  √  |                |               |      |   39 (A100)   |      1.2      |
+|  3  |     |       √        |               |      |   33 (A100)   |      1.1      |
+|  4  |  √  |       √        |               |      |   25 (A100)   |      1.3      |
+|  5  |     |       √        |       √       |      |      18       |      2.2      |
+|  6  |  √  |       √        |       √       |      |      13       |      1.6      |
+|  7  |     |       √        |       √       |  √   |      14       |      2.9      |
+|  8  |  √  |       √        |       √       |  √   |      8.5      |      2.4      |
+
+**6. Support for the [V3Det](configs/v3det/README.md) dataset, a large-scale detection dataset with over 13,000 categories.**
+
+<div align=center>
+    <img width=960 src="https://github.com/open-mmlab/mmdetection/assets/17425982/9c216387-02be-46e6-b0f2-b856f80f6d84"/>
+</div>
+
 We are excited to announce our latest work on real-time object recognition tasks, **RTMDet**, a family of fully convolutional single-stage detectors. RTMDet not only achieves the best parameter-accuracy trade-off on object detection from tiny to extra-large model sizes but also obtains new state-of-the-art performance on instance segmentation and rotated object detection tasks. Details can be found in the [technical report](https://arxiv.org/abs/2212.07784). Pre-trained models are [here](configs/rtmdet).
 
 [![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/rtmdet-an-empirical-study-of-designing-real/real-time-instance-segmentation-on-mscoco)](https://paperswithcode.com/sota/real-time-instance-segmentation-on-mscoco?p=rtmdet-an-empirical-study-of-designing-real)
@@ -119,13 +165,6 @@ We are excited to announce our latest work on real-time object recognition tasks
 <img src="https://user-images.githubusercontent.com/12907710/208044554-1e8de6b5-48d8-44e4-a7b5-75076c7ebb71.png"/>
 </div>
 
-**v3.1.0** was released in 30/6/2023:
-
-- Supports tracking algorithms including multi-object tracking (MOT) algorithms SORT, DeepSORT, StrongSORT, OCSORT, ByteTrack, QDTrack, and video instance segmentation (VIS) algorithm MaskTrackRCNN, Mask2Former-VIS.
-- Support [ViTDet](projects/ViTDet)
-- Supports inference and evaluation of multimodal algorithms [GLIP](configs/glip) and [XDecoder](projects/XDecoder), and also supports datasets such as COCO semantic segmentation, COCO Caption, ADE20k general segmentation, and RefCOCO. GLIP fine-tuning will be supported in the future.
-- Provides a [gradio demo](https://github.com/open-mmlab/mmdetection/blob/dev-3.x/projects/gradio_demo/README.md) for image type tasks of MMDetection, making it easy for users to experience.
-
 ## Installation
 
 Please refer to [Installation](https://mmdetection.readthedocs.io/en/latest/get_started.html) for installation instructions.

diff --git a/README_zh-CN.md b/README_zh-CN.md
@@ -102,6 +102,53 @@ MMDetection 是一个基于 PyTorch 的目标检测开源工具箱。它是 [Ope
 
 ### 亮点
 
+**v3.2.0** 版本已经在 2023.10.12 发布：
+
+**1. 检测 Transformer SOTA 模型大合集**
+(1) 支持了 [DDQ](configs/ddq/README.md)、[CO-DETR](projects/CO-DETR/README.md)、[AlignDETR](projects/AlignDETR/README.md) 和 [H-DINO](projects/HDINO/README.md) 4 个更新更强的 SOTA Transformer 模型
+(2) 基于 CO-DETR,  MMDet 中发布了 COCO 性能为 64.1 mAP 的模型
+(3) DINO 等算法支持 AMP/Checkpoint/FrozenBN，可以有效降低显存
+
+**2. [提供了全面的 CNN 和 Transformer 的性能对比](projects/RF100-Benchmark/README_zh-CN.md)**
+RF100 是由 100 个现实收集的数据集组成，包括 7 个域，可以验证 DINO 等 Transformer 模型和 CNN 类算法在不同场景不同数据量下的性能差异。用户可以用这个 Benchmark 快速验证自己的算法在不同场景下的鲁棒性。
+
+<div align=center>
+<img src="https://github.com/open-mmlab/mmdetection/assets/17425982/86420903-36a8-410d-9251-4304b9704f7d"/>
+</div>
+
+**3. 支持了 [GLIP](configs/glip/README.md) 和 [Grounding DINO](configs/grounding_dino/README.md) 微调，全网唯一支持 Grounding DINO 微调**
+MMDet 中的 Grounding DINO 是全网唯一支持微调的算法库，且性能高于官方 1 个点，当然 GLIP 也比官方高。
+我们还提供了详细的 Grounding DINO 在自定义数据集上训练评估的流程，欢迎大家试用。
+
+|       Model        | Backbone |   Style   |  COCO mAP  | Official COCO mAP |
+| :----------------: | :------: | :-------: | :--------: | :---------------: |
+|  Grounding DINO-T  |  Swin-T  | Zero-shot |    48.5    |       48.4        |
+|  Grounding DINO-T  |  Swin-T  | Finetune  | 58.1(+0.9) |       57.2        |
+|  Grounding DINO-B  |  Swin-B  | Zero-shot |    56.9    |       56.7        |
+|  Grounding DINO-B  |  Swin-B  | Finetune  |    59.7    |                   |
+| Grounding DINO-R50 |   R50    |  Scratch  | 48.9(+0.8) |       48.1        |
+
+**4. 支持开放词汇检测算法 [Detic](projects/Detic_new/README.md) 并提供多数据集联合训练可能**
+
+**5. 轻松使用 [FSDP 和 DeepSpeed 训练检测模型](projects/example_largemodel/README_zh-CN.md)**
+
+| ID  | AMP | GC of Backbone | GC of Encoder | FSDP | Peak Mem (GB) | Iter Time (s) |
+| :-: | :-: | :------------: | :-----------: | :--: | :-----------: | :-----------: |
+|  1  |     |                |               |      |   49 (A100)   |      0.9      |
+|  2  |  √  |                |               |      |   39 (A100)   |      1.2      |
+|  3  |     |       √        |               |      |   33 (A100)   |      1.1      |
+|  4  |  √  |       √        |               |      |   25 (A100)   |      1.3      |
+|  5  |     |       √        |       √       |      |      18       |      2.2      |
+|  6  |  √  |       √        |       √       |      |      13       |      1.6      |
+|  7  |     |       √        |       √       |  √   |      14       |      2.9      |
+|  8  |  √  |       √        |       √       |  √   |      8.5      |      2.4      |
+
+**6. 支持了 [V3Det](configs/v3det/README.md) 1.3w+ 类别的超大词汇检测数据集**
+
+<div align=center>
+    <img width=960 src="https://github.com/open-mmlab/mmdetection/assets/17425982/9c216387-02be-46e6-b0f2-b856f80f6d84"/>
+</div>
+
 我们很高兴向大家介绍我们在实时目标识别任务方面的最新成果 RTMDet，包含了一系列的全卷积单阶段检测模型。 RTMDet 不仅在从 tiny 到 extra-large 尺寸的目标检测模型上实现了最佳的参数量和精度的平衡，而且在实时实例分割和旋转目标检测任务上取得了最先进的成果。 更多细节请参阅[技术报告](https://arxiv.org/abs/2212.07784)。 预训练模型可以在[这里](configs/rtmdet)找到。
 
 [![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/rtmdet-an-empirical-study-of-designing-real/real-time-instance-segmentation-on-mscoco)](https://paperswithcode.com/sota/real-time-instance-segmentation-on-mscoco?p=rtmdet-an-empirical-study-of-designing-real)
@@ -118,13 +165,6 @@ MMDetection 是一个基于 PyTorch 的目标检测开源工具箱。它是 [Ope
 <img src="https://user-images.githubusercontent.com/12907710/208044554-1e8de6b5-48d8-44e4-a7b5-75076c7ebb71.png"/>
 </div>
 
-**v3.1.0** 版本已经在 2023.6.30 发布：
-
-- 支持 Tracking 类算法，包括多目标跟踪 MOT 算法 SORT、DeepSORT、StrongSORT、OCSORT、ByteTrack、QDTrack 和视频实例分割 VIS 算法 MaskTrackRCNN、Mask2Former-VIS。
-- 支持 [ViTDet](projects/ViTDet)
-- 支持多模态开放检测算法 [GLIP](configs/glip) 和 [XDecoder](projects/XDecoder) 推理和评估，并同时支持了 COCO 语义分割、COCO Caption、ADE20k 通用分割、RefCOCO 等数据集。后续将支持 GLIP 微调
-- 提供了包括 MMDetection 图片任务的 [gradio demo](https://github.com/open-mmlab/mmdetection/blob/dev-3.x/projects/gradio_demo/README.md)，方便用户快速体验
-
 ## 安装
 
 请参考[快速入门文档](https://mmdetection.readthedocs.io/zh_CN/latest/get_started.html)进行安装。

diff --git a/docker/serve/Dockerfile b/docker/serve/Dockerfile
@@ -4,7 +4,7 @@ ARG CUDNN="8"
 FROM pytorch/pytorch:${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel
 
 ARG MMCV="2.0.0rc4"
-ARG MMDET="3.1.0"
+ARG MMDET="3.2.0"
 
 ENV PYTHONUNBUFFERED TRUE
 

diff --git a/docker/serve_cn/Dockerfile b/docker/serve_cn/Dockerfile
@@ -4,7 +4,7 @@ ARG CUDNN="8"
 FROM pytorch/pytorch:${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel
 
 ARG MMCV="2.0.0rc4"
-ARG MMDET="3.1.0"
+ARG MMDET="3.2.0"
 
 ENV PYTHONUNBUFFERED TRUE
 

diff --git a/docs/en/notes/changelog.md b/docs/en/notes/changelog.md
@@ -1,5 +1,73 @@
 # Changelog of v3.x
 
+## v3.1.0 (12/10/2023)
+
+### Highlights
+
+**(1) Detection Transformer SOTA Model Collection**
+
+- Supported four updated and stronger SOTA Transformer models: DDQ, CO-DETR, AlignDETR, and H-DINO.
+- Based on CO-DETR, MMDet released a model with a COCO performance of 64.1 mAP.
+- Algorithms such as DINO support AMP/Checkpoint/FrozenBN, which can effectively reduce memory usage.
+
+**(2) Comprehensive Performance Comparison between CNN and Transformer**
+
+RF100 consists of a dataset collection of 100 real-world datasets, including 7 domains. It can be used to assess the performance differences of Transformer models like DINO and CNN-based algorithms under different scenarios and data volumes. Users can utilize this benchmark to quickly evaluate the robustness of their algorithms in various scenarios.
+
+**(3) Support for GLIP and Grounding DINO fine-tuning, the only algorithm library that supports Grounding DINO fine-tuning**
+
+The Grounding DINO algorithm in MMDet is the only library that supports fine-tuning. Its performance is one point higher than the official version, and of course, GLIP also outperforms the official version.
+We also provide a detailed process for training and evaluating Grounding DINO on custom datasets. Everyone is welcome to give it a try.
+
+**(4) Support for the open-vocabulary detection algorithm Detic and multi-dataset joint training.**
+
+**(5) Training detection models using FSDP and DeepSpeed.**
+
+**(6) Support for the V3Det dataset, a large-scale detection dataset with over 13,000 categories.**
+
+### New Features
+
+- Support CO-DETR/DDQ/AlignDETR/H-DINO
+- Support GLIP and Grounding DINO fine-tuning
+- Support Detic and Multi-Datasets training (#10926)
+- Support V3Det and benchmark (#10938)
+- Support Roboflow 100 Benchmark (#10915)
+- Add custom dataset of grounding dino (#11012)
+- Release RTMDet-X p6 (#10993)
+- Support AMP of DINO (#10827)
+- Support FrozenBN (#10845)
+- Add new configuration files for `QDTrack/DETR/RTMDet/MaskRCNN/DINO/DeformableDETR/MaskFormer` algorithm
+- Add a new script to support the WBF (#10808)
+- Add `large_image_demo` (#10719)
+- Support download dataset from OpenXLab (#10799)
+- Update to support torch2onnx for DETR series models (#10910)
+- Translation into Chinese of an English document (#10744, #10756, #10805, #10848)
+
+### Bug Fixes
+
+- Fix name error in DETR metafile.yml (#10595)
+- Fix device of the tensors in `set_nms` (#10574)
+- Remove some unicode chars from `en/` docs (#10648)
+- Fix download dataset with mim script. (#10727)
+- Fix export to torchserve (#10694)
+- Fix typo in `mask-rcnn_r50_fpn_1x-wandb_coco` (#10757)
+- Fix `eval_recalls` error in `voc_metric` (#10770)
+- Fix torch version comparison (#10934)
+- Fix incorrect behavior to access train pipeline from ConcatDataset in `analyze_results.py` (#11004)
+
+### Improvements
+
+- Update `useful_tools.md` (#10587)
+- Update Instance segmentation Tutorial (#10711)
+- Update `train.py` to compat with new config (#11025)
+- Support `torch2onnx` for maskformer series (#10782)
+
+### Contributors
+
+A total of 36 developers contributed to this release.
+
+Thank @YQisme, @nskostas, @max-unfinity, @evdcush, @Xiangxu-0103, @ZhaoCake, @RangeKing, @captainIT, @ODAncona, @aaronzs, @zeyuanyin, @gotjd709, @Musiyuan, @YanxingLiu, @RunningLeon, @ytzfhqs, @zhangzhidaSunny, @yeungkong, @crazysteeaam, @timerring, @okotaku, @apatsekin, @Morty-Xu, @Markson-Young, @ZhaoQiiii, @Kuro96, @PhoenixZ810, @yhcao6, @myownskyW7, @jiongjiongli, @Johnson-Wang, @ryylcc, @guyleaf, @agpeshal, @SimonGuoNjust, @hhaAndroid
+
 ## v3.1.0 (30/6/2023)
 
 ### Highlights

diff --git a/docs/en/notes/faq.md b/docs/en/notes/faq.md
@@ -46,7 +46,8 @@ Compatible MMDetection, MMEngine, and MMCV versions are shown as below. Please c
 
 | MMDetection version |      MMCV version       |     MMEngine version     |
 | :-----------------: | :---------------------: | :----------------------: |
-|        main         |  mmcv>=2.0.0, \<2.1.0   | mmengine>=0.7.1, \<1.0.0 |
+|        main         |  mmcv>=2.0.0, \<2.2.0   | mmengine>=0.7.1, \<1.0.0 |
+|        3.2.0        |  mmcv>=2.0.0, \<2.2.0   | mmengine>=0.7.1, \<1.0.0 |
 |        3.1.0        |  mmcv>=2.0.0, \<2.1.0   | mmengine>=0.7.1, \<1.0.0 |
 |        3.0.0        |  mmcv>=2.0.0, \<2.1.0   | mmengine>=0.7.1, \<1.0.0 |
 |      3.0.0rc6       | mmcv>=2.0.0rc4, \<2.1.0 | mmengine>=0.6.0, \<1.0.0 |

diff --git a/docs/zh_cn/notes/faq.md b/docs/zh_cn/notes/faq.md
@@ -46,7 +46,8 @@ export DYNAMO_CACHE_SIZE_LIMIT = 4
 
   | MMDetection 版本 |        MMCV 版本        |      MMEngine 版本       |
   | :--------------: | :---------------------: | :----------------------: |
-  |       main       |  mmcv>=2.0.0, \<2.1.0   | mmengine>=0.7.1, \<1.0.0 |
+  |       main       |  mmcv>=2.0.0, \<2.2.0   | mmengine>=0.7.1, \<1.0.0 |
+  |      3.2.0       |  mmcv>=2.0.0, \<2.2.0   | mmengine>=0.7.1, \<1.0.0 |
   |      3.1.0       |  mmcv>=2.0.0, \<2.1.0   | mmengine>=0.7.1, \<1.0.0 |
   |      3.0.0       |  mmcv>=2.0.0, \<2.1.0   | mmengine>=0.7.1, \<1.0.0 |
   |     3.0.0rc6     | mmcv>=2.0.0rc4, \<2.1.0 | mmengine>=0.6.0, \<1.0.0 |

diff --git a/mmdet/__init__.py b/mmdet/__init__.py
@@ -6,7 +6,7 @@
 from .version import __version__, version_info
 
 mmcv_minimum_version = '2.0.0rc4'
-mmcv_maximum_version = '3.0.0'
+mmcv_maximum_version = '2.2.0'
 mmcv_version = digit_version(mmcv.__version__)
 
 mmengine_minimum_version = '0.7.1'

diff --git a/requirements/mminstall.txt b/requirements/mminstall.txt
@@ -1,2 +1,2 @@
-mmcv>=2.0.0rc4,<3.0.0
+mmcv>=2.0.0rc4,<2.2.0
 mmengine>=0.7.1,<1.0.0
diff --git a/requirements/readthedocs.txt b/requirements/readthedocs.txt
@@ -1,4 +1,4 @@
-mmcv>=2.0.0rc4,<2.1.0
+mmcv>=2.0.0rc4,<2.2.0
 mmengine>=0.7.1,<1.0.0
 scipy
 torch