https://github.com/tinyvision/solider-humanpose
https://github.com/tinyvision/solider-humanpose
Last synced: 5 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/tinyvision/solider-humanpose
- Owner: tinyvision
- License: apache-2.0
- Created: 2023-07-21T08:16:14.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2023-07-21T08:19:56.000Z (over 2 years ago)
- Last Synced: 2024-12-23T16:45:21.983Z (over 1 year ago)
- Language: Python
- Size: 22.7 MB
- Stars: 3
- Watchers: 1
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- Contributing: .github/CONTRIBUTING.md
- License: LICENSE
- Code of conduct: .github/CODE_OF_CONDUCT.md
- Citation: CITATION.cff
Awesome Lists containing this project
README
# SOLIDER on [Human Pose]
This repo provides details about how to use [SOLIDER](https://github.com/tinyvision/SOLIDER) pretrained representation on human parsing task.
We modify the code from [mmpose](https://github.com/open-mmlab/mmpose), and you can refer to the original repo for more details.
## Installation and Datasets
Details of installation and dataset preparation can be found in [mmpose-installation](https://mmpose.readthedocs.io/en/latest/installation.html).
## Prepare Pre-trained Models
Step 1. Download models from [SOLIDER](https://github.com/tinyvision/SOLIDER), or use [SOLIDER](https://github.com/tinyvision/SOLIDER) to train your own models.
Steo 2. Put the pretrained models under the `pretrained` file, and rename their names as `./pretrained/solider_swin_tiny(small/base).pth`
## Training
Train with single GPU or multiple GPUs:
```shell
sh run_train.sh
```
## Performance
| Method | Model | COCO(AP/AR) |
| ------ | :---: | :---: |
| SOLIDER | Swin Tiny | 74.4/79.6 |
| SOLIDER | Swin Small | 76.3/81.3 |
| SOLIDER | Swin Base | 76.6/81.5 |
- We use the pretrained models from [SOLIDER](https://github.com/tinyvision/SOLIDER).
- The semantic weight we used in these experiments is 0.8.
## Citation
If you find this code useful for your research, please cite our paper
```
@inproceedings{chen2023beyond,
title={Beyond Appearance: a Semantic Controllable Self-Supervised Learning Framework for Human-Centric Visual Tasks},
author={Weihua Chen and Xianzhe Xu and Jian Jia and Hao Luo and Yaohua Wang and Fan Wang and Rong Jin and Xiuyu Sun},
booktitle={The IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2023},
}
```