A repo containing codes and data for fashion image captioning for paper Fashion Captioning: Towards Generating Accurate Descriptions with Semantic Rewards. Code and Data.
If you have difficulty running the training scripts in tools. You can try installing this repo as a python package:
python -m pip install -e .
| Model | B-1 | B-2 | B-3 | B-4 | METEOR | ROUGE-L | CIDEr | SPICE | mAP | ACC |
|---|---|---|---|---|---|---|---|---|---|---|
| NewFC | 19.9 | 8.1 | 3.8 | 2.2 | 7.5 | 16.8 | 21.1 | 6.9 | 0 | 0 |
| SAT | 24.4 | 11.4 | 6.4 | 4.3 | 9.5 | 19.9 | 35.2 | 9.8 | -- | -- |
| BUTD | 24.7 | 11.5 | 6.5 | 4.4 | 9.7 | 19.6 | 36.9 | 10.1 | -- | -- |
| Trans | 24.4 | 11.7 | 6.2 | 4.2 | 10.2 | 19.9 | 36.7 | 9.9 | -- | -- |
| SAT+Cls | 25.7 | 12.5 | 7.5 | 5.3 | 10.2 | 20.6 | 43.2 | 11.1 | -- | -- |
- The max_length should be set at 30 instead of 20, as over 80% of the captions having number of words <= 30.
- The dataset is under license in the LICENSE file.
- No commercial use.
If you use this data, please cite:
@inproceedings{XuewenECCV20Fashion,
Author = {Xuewen Yang and Heming Zhang and Di Jin and Yingru Liu and Chi-Hao Wu and Jianchao Tan and Dongliang Xie and Jue Wang and Xin Wang},
Title = {Fashion Captioning: Towards Generating Accurate Descriptions with Semantic Rewards},
booktitle = {ECCV},
Year = {2020}
}
or
@ARTICLE{2020arXiv200802693Y,
author = {{Yang}, Xuewen and {Zhang}, Heming and {Jin}, Di and {Liu}, Yingru and {Wu}, Chi-Hao and {Tan}, Jianchao and {Xie}, Dongliang and {Wang}, Jue and {Wang}, Xin},
title = "{Fashion Captioning: Towards Generating Accurate Descriptions with Semantic Rewards}",
journal = {arXiv e-prints},
year = 2020,
eprint = {2008.02693},
adsurl = {https://ui.adsabs.harvard.edu/abs/2020arXiv200802693Y},
}