fouriye / dcrgan-tmm

DCRGAN-TMM

This is a PyTorch implementation for the paper "Disentangling Semantic-to-visual Confusion for Zero-shot Learning" in IEEE Transactions on Multimedia 2021.

Prepare dataset

Download the data and uncompress it to the folder '../data'.

Reproduce results

Run demo.sh for CUB, AWA1, APY, SUN dataset

bash demo.sh

Citation

If you think this code is useful in your research or wish to refer to the baseline results published in our paper, please use the following BibTeX entry.

@ARTICLE{ye2021disentangling,
  author={Ye, Zihan and Hu, Fuyuan and Lyu, Fan and Li, Linyan and Huang, Kaizhu},
  journal={IEEE Transactions on Multimedia}, 
  title={Disentangling Semantic-to-Visual Confusion for Zero-Shot Learning}, 
  year={2022},
  volume={24},
  number={},
  pages={2828-2840},
  doi={10.1109/TMM.2021.3089017}}

dcrgan-tmm's Issues

model_SRN_Iter_20000.tar

Sorry to bother you, when training DCR_GAN, do all datasets select the SRN obtained by 20000 iterations in pre-training, i.e. model_SRN_Iter_20000.tar?