This repo provides details about how to use SOLIDER pretrained representation on attribute recognition task. We modify the code from Rethinking_of_PAR, and you can refer to the original repo for more details.
Details of installation and dataset preparation can be found in Rethinking_of_PAR.
Step 1. Download models from SOLIDER, or use 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
Train with single GPU or multiple GPUs:
sh run.sh
Method | Model | PETA_ZS (mA) |
RAP_ZS (mA) |
PA100K (mA) |
---|---|---|---|---|
SOLIDER | Swin Tiny | 74.37 | 74.23 | 84.14 |
SOLIDER | Swin Small | 76.21 | 75.95 | 86.25 |
SOLIDER | Swin Base | 76.43 | 76.42 | 86.37 |
- We use the pretrained models from SOLIDER.
- The semantic weight is set to 0.8 in these experiments.
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},
}