Checkpoints, logs and source code for AAAI-23 paper 'Data-Efficient Image Quality Assessment with Attention-Panel Decoder'

• [06/2023] We released the source code of 'DEIQT', check the code on GitHub
• [04/2023] We released our work 'DEIQT', the paper is now on Arxiv.

• Checkpoints & Logs
• Initialization

• Python
• PyTorch

Download the Pre-trained ViT-S[224] weights from DeiT III

This project use WandB to log information and report. Remember to adjust the code in main.py to suit your research.

CUDA_VISIBLE_DEVICES=0,1,2,3 OMP_NUM_THREADS=1 torchrun --nnodes 1 --nproc_per_node 4 --master_port 49935  main.py \
--cfg [CONFIG_PATH] \
--data-path [YOUR_DATA_PATH] \
--output [LOG_PATH] \
--tag [REMARK_TAG] \
--repeat \
--rnum [TARGET_REPEAT_NUM]

If you find this project helpful in your research, please consider citing our papers:

@inproceedings{qin2023deiqt,
  title={Data-Efficient Image Quality Assessment with Attention-Panel Decoder},
  author={Guanyi Qin and Runze Hu and Yutao Liu and Xiawu Zheng and Haotian Liu and Xiu Li and Yan Zhang},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)},
  year={2023}
}

We borrowed some parts from the following open-source projects:

• HyperIQA
• Swin-Transformer
• DeiT III

Many thanks to them.