https://github.com/visionxlab/pwood
https://github.com/visionxlab/pwood
Last synced: 10 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/visionxlab/pwood
- Owner: VisionXLab
- Created: 2025-03-13T06:29:21.000Z (about 1 year ago)
- Default Branch: HBox
- Last Pushed: 2025-07-04T06:17:34.000Z (10 months ago)
- Last Synced: 2025-07-04T06:33:42.348Z (10 months ago)
- Language: Python
- Size: 11 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Partial Weakly-Supervised Oriented Object Detection
Mingxin Liu
Peiyuan Zhang
Yuan Liu
Wei Zhang
Yue Zhou
Ning Liao
Ziyang Gong
Junwei Luo
Zhirui Wang
Yi Yu
Xue Yang
## Introduction
We propose the first Partial Weakly-Supervised Oriented Object Detection (PWOOD) framework based on partially weak annotations (horizontal boxes or single points), which can efficiently leverage large amounts of unlabeled data, significantly outperforming weakly supervised algorithms trained with partially weak annotations, and also offers a lower cost solution.
## Installation
``` shell
conda create -n mm python==3.8 -y
conda activate mm
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116
pip install -U openmim
mim install mmcv-full
mim install mmdet\<3.0.0
pip install scikit-learn
pip install prettytable
# For Point branch
cd mmrotate
pip install -v -e .
```
## Data Preparation
### DOTA
#### 1. Labeled/Unlabeled Data Division
To divide the DOTA- v1.0/v1.5 dataset into labeled and unlabeled data, please refer to [Data preparation of SOOD
](https://github.com/HamPerdredes/SOOD).
To divide the DOTA- v2.0 into labeled and unlabeled data, please refer to [data_list/dotav2](https://github.com/123sio/PWOOD/tree/HBox/data_list/dotav2).
#### 2. Data Split
For details on how to split the DOTA dataset into patches, please refer to the [official implementation](https://github.com/open-mmlab/mmrotate/blob/main/tools/data/dota/README.md) .
After split, the data folder should be organized as follows:
```
split_ss_dota_vxx
├── train
│ ├── images
│ └── annfiles
├── val
│ ├── images
│ └── annfiles
├── train_xx_labeled
│ ├── images
│ └── annfiles
└──train_xx_unlabeled
├── images
└── annfiles
```
### DIOR
To divide the DIOR into labeled and unlabeled data, please refer to [data_list/dior](https://github.com/123sio/PWOOD/tree/HBox/data_list/dior).
## Train
```bash
#2 GPU
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nnodes=1 \
--node_rank=0 --master_addr="127.0.0.1" --nproc_per_node=2 --master_port=25510 \
train.py configs_dota15/xxx/xxx.py \
--launcher pytorch \
--work-dir work_dir/xxx/
```
## Test
```bash
python test.py configs_dota15/xxx/xxx.py work_dir/xxx/xxx.pth
```
## Weight
### DOTA- v1.0
Labeled Data | mAP | Config | Model | Log |
| :-----------: | :--: |:-----: | :----: | :-----:|
| 20% | 62.93 | [semi_h2rv2_adamw_dotav1_20p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dotav1/semi_h2rv2_adamw_dotav1_20p.py) | [best_0.629314_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dota1_0/20p/best_0.629314_mAP.pth) | [dotav1_20p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dota1_0/20p/20250303_111353.log.json) |
| 30% | 65.42 | [semi_h2rv2_adamw_dotav1_30p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dotav1/semi_h2rv2_adamw_dotav1_30p.py) | [best_0.654153_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dota1_0/30p/best_0.654153_mAP.pth) | [dotav1_30p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dota1_0/30p/20250310_193742.log.json) |
### DOTA- v1.5
Labeled Data | mAP | Config | Model | Log |
| :-----------: | :--: |:-----: | :----: | :-----:|
| 10% | 52.87 | [semi_h2rv2_adamw_dota15_10p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/semi_h2rv2_adamw_dota15_10p.py) | [best_0.528748_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/gmm/10p_lr/best_0.528748_mAP.pth) | [dotav15_10p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/gmm/10p_lr/20250219_224359.log.json) |
| 20% | 59.36 | [semi_h2rv2_adamw_dota15_20p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/semi_h2rv2_adamw_dota15_20p.py) | [best_0.593614_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/gmm/best_0.593614_mAP.pth) | [dotav15_20p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/gmm/20250217_202030.log.json) |
| 30% | 61.58 | [semi_h2rv2_adamw_dota15_30p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/semi_h2rv2_adamw_dota15_30p.py) | [best_0.615836_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/gmm/30p_lr/best_0.615836_mAP.pth) | [dotav15_30p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/gmm/30p_lr/20250219_223950.log.json) |
### DOTA- v2.0
Labeled Data | mAP | Config | Model | Log |
| :-----------: | :--: |:-----: | :----: | :-----:|
| 10% | 31.30 | [semi_h2rv2_adamw_dota2_10p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dotav2/semi_h2rv2_adamw_dota2_10p.py)| [best_0.310266_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dotav2/10p/best_0.310266_mAP.pth)| [dotav2_10p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dotav2/10p/20250313_224047.log.json)|
| 20% | 36.39 | [semi_h2rv2_adamw_dota2_20p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dotav2/semi_h2rv2_adamw_dota2_20p.py)| [best_0.363926_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dotav2/best_0.363926_mAP.pth) | [dotav2_20p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dotav2/20250304_174131.log.json) |
| 30% | 40.27 | [semi_h2rv2_adamw_dota2_30p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dotav2/semi_h2rv2_adamw_dota2_30p.py)| [best_0.402659_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dotav2/30p/pro_data/best_0.402659_mAP.pth) | [dotav2_30p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dotav2/30p/pro_data/20250321_142715.log.json) |
### DIOR
Labeled Data | mAP | Config | Model | Log |
| :-----------: | :--: |:-----: | :----: | :-----:|
| 10% | 54.33 | [semi_h2rv2_adamw_dior_10p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dior/semi_h2rv2_adamw_dior_10p.py) | [best_0.543296_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dior/gmm/10p/best_0.543296_mAP.pth) | [doir_10p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dior/gmm/10p/20250227_202752.log.json) |
| 20% | 57.89 | [semi_h2rv2_adamw_dior_20p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dior/semi_h2rv2_adamw_dior_20p.py) | [best_0.578923_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dior/gmm/20p/best_0.578923_mAP.pth) | [dior_20p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dior/gmm/20p/20250227_205200.log.json) |
| 30% | 60.42 | [semi_h2rv2_adamw_dior_30p.py](https://github.com/123sio/PWOOD/blob/HBox/configs_dota15/pwood/dior/semi_h2rv2_adamw_dior_30p.py) | [best_0.604248_mAP.pth](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dior/gmm/30p/best_0.604248_mAP.pth) | [dior_30p_log](https://huggingface.co/Xm4nQ8/weight/blob/main/work_dir_h/PWOOD/dior/gmm/30p/20250301_071406.log.json) |
## Guide
If you need Point version, please switch to the [Point branch](https://github.com/123sio/PWOOD/tree/Point).
## Citation
```bibtex
@misc{liu2025partialweaklysupervisedorientedobject,
title={Partial Weakly-Supervised Oriented Object Detection},
author={Mingxin Liu and Peiyuan Zhang and Yuan Liu and Wei Zhang and Yue Zhou and Ning Liao and Ziyang Gong and Junwei Luo and Zhirui Wang and Yi Yu and Xue Yang},
year={2025},
eprint={2507.02751},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2507.02751},
}
```