DM-Adapter

KO | EN

gitlite — search

#python #java #python3 #arduino #golang #machine-learning #rust #html #flask #javascript #seismology #nodejs

★ 17 Open GitHub ↗

About [AAAI 2025] Official repository of paper titled "DM-Adapter: Domain-Aware Mixture-of-Adapters for Text-Based Person Retrieval"

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# DM-Adapter: Domain-Aware Mixture-of-Adapters for Text-Based Person Retrieval

## Highlights

The goal of this work is to enhance implicit fine-grained knowledge transferring,  offering the best trade-off between performance and parameter efficiency.

## Usage
### Requirements
we use single NVIDIA 4090 24G GPU for training and evaluation. 
```
pytorch 1.12.1
torchvision 0.13.1
prettytable
easydict
```

### Prepare Datasets
Download the CUHK-PEDES dataset from [here](https://github.com/ShuangLI59/Person-Search-with-Natural-Language-Description), ICFG-PEDES dataset from [here](https://github.com/zifyloo/SSAN) and RSTPReid dataset form [here](https://github.com/NjtechCVLab/RSTPReid-Dataset)

## Training

```python
python train.py \
--name baseline \
--img_aug \
--batch_size 128 \
--MLM \
--dataset_name $DATASET_NAME \
--loss_names 'sdm+aux' \
--num_epoch 60 \
--root_dir '.../dataset_reid' \
--lr 3e-4 \
--num_experts 6 \
--topk 2 \
--reduction 8
```

## Testing

```python
python test.py --config_file 'path/to/model_dir/configs.yaml'
```

## Citation
If you find this code useful for your research, please cite our paper.

```tex
@inproceedings{liu2025DM-Adapter,
  title={DM-Adapter: Domain-Aware Mixture-of-Adapters for Text-Based Person Retrieval},
  author={Liu, Yating and Liu, Zimo and Lan, Xiangyuan and Yang, Wenming and Li, Yaowei and Liao, Qingmin},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2025}
}
```

## Acknowledgments
Some components of this code implementation are adopted from [CLIP](https://github.com/openai/CLIP), [IRRA](https://github.com/anosorae/IRRA) and [Mixtral](https://github.com/mistralai/mistral-inference). We sincerely appreciate for their contributions.

DM-Adapter

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README