Skip to content

Latest commit

 

History

History
53 lines (32 loc) · 1.83 KB

README.md

File metadata and controls

53 lines (32 loc) · 1.83 KB
Raw

Attribute Attention for Semantic Disambiguation in Zero-Shot Learning

This repository contains the public release of the Python implementation of

Attribute Attention for Semantic Disambiguation in Zero-Shot Learning

Yang Liu, Jishun Guo, Deng Cai, Xiaofei He.

framework

If you use this code or find this work useful for your research, please cite:

@inproceedings{Liu_2019_ICCV,
  title={Attribute Attention for Semantic Disambiguation in Zero-Shot Learning},
  author={Liu, Yang and Guo, Jishun and Cai, Deng and He, Xiaofei},
  booktitle={The IEEE International Conference on Computer Vision (ICCV)},
  month={Oct},
  year={2019}
}

Performance

cZSL

czsl

gZSL

gzsl

Start Up

Implemented and tested on Ubuntu 16.04 with Python 3.6 and Pytorch 1.0.1. Experiments are conducted on AwA2, CUB and SUN datasets.

We use AwA2 file format as default detailed in ./data/ folder and images should be downloaded and renamed as ./data/*/JPEGImages. It is important to note that several cusomization work should be done for SUN dataset to maintain the same file format.

Basic Usage

Train

Use experiments/run_trainer.py to train the network. Run help to view all the possible parameters. We provide several config files under ./configs/ folder. Example usage:

python experiments/run_trainer.py --cfg ./configs/self_adaptation/VGG19_AwA2_PS_C.yaml

Feel free to download the reported checkpoints.

Test

Use experiments/run_evaluator.py to evaluate the network with self_adaptation and experiments/run_evaluator_hybrid.py with hybrid method.