CANZSL: Cycle-Consistent Adversarial Networks for Zero-Shot Learning from Natural Language

code for the paper CANZSL: Cycle-Consistent Adversarial Networks for Zero-Shot Learning from Natural Language WACV 2020.

The code is based on the implementation of GAZSL [1].

Data: You can download the dataset CUBird and NABird
Put the uncompressed data to the folder "data"

Reproduce results

CUBird SCS mode && SCE mode

python train.py --dataset CUB2011 --splitmode easy
python train.py --dataset CUB2011 --splitmode hard

NABird SCS mode && SCE mode

python train.py --dataset NABird --splitmode easy
python train.py --dataset NABird --splitmode hard

Reference:

[1] Yizhe Zhu, Mohamed Elhoseiny, Bingchen Liu, Xi Peng and Ahmed Elgammal "A Generative Adversarial Approach for Zero-Shot Learning from Noisy Texts", CVPR, 2018