ZSL_GAN

code for the conference and the journal versions of the paper:

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

Processed feature Data:

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

Raw wikipedia article data:

Raw wikipedia article data of CUBird and NABird, as well as detailed merging information of NABird, can be obtained here.

Reproduce results

CUBird SCS mode && SCE mode

python train_CUB.py --splitmode easy
python train_CUB.py --splitmode hard

NABird SCS mode && SCE mode

python train_NAB.py --splitmode easy
python train_NAB.py --splitmode hard

Results evaluated on GBU setting[1]

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

python train_GBU.py --dataset CUB1 --preprocessing --z_dim 100
python train_GBU.py --dataset AWA1 --preprocessing --z_dim 10
python train_GBU.py --dataset AWA2 --preprocessing --z_dim 10
python train_GBU.py --dataset APY --preprocessing --z_dim 10
python train_GBU.py --dataset SUN --preprocessing --z_dim 10
python train_GBU.py --dataset FLO --preprocessing --z_dim 100

Note: The results of work [2-11] are copied from TABLE6 in [1]. The results of work [12] are obtained from the authors' report. Thanks to Li Zhang for the template of README.

Citation

If you find this implementation or the analysis conducted in our report helpful, please consider citing:

@inproceedings{Yizhe_ZSL_2018,  
    Author = {Yizhe Zhu, Mohamed Elhoseiny, Bingchen Liu, Xi Peng and Ahmed Elgammal},  
    Title = {A Generative Adversarial  Approach for Zero-Shot Learning from Noisy Texts},  
    Booktitle = {CVPR},  
    Year = {2018}  
}

TODO:

• Add result evaluated on GBU setting

Update:

• 2022/01/10 update the link of shared raw wiki data
• 2018/10/22 add AUC Score for CUBird and NABird
• 2018/07/23 add results on GBU setting
• 2018/07/18 add the experiments of NABird.
• 2018/07/18 merge the train.py and test.py to one file
• 2018/07/17 make code compatable with python2&3

References

• [1] Zero-Shot Learning - A Comprehensive Evaluation of the Good, the Bad and the Ugly. Yongqin Xian, Christoph H. Lampert, Bernt Schiele, Zeynep Akata. arXiv, 2017.
• [2] Attribute-Based Classification forZero-Shot Visual Object Categorization. Christoph H. Lampert, Hannes Nickisch and Stefan Harmeling. PAMI, 2014.
• [3] Zero-Shot Learning by Convex Combination of Semantic Embeddings. Mohammad Norouzi, Tomas Mikolov, Samy Bengio, Yoram Singer, Jonathon Shlens, Andrea Frome, Greg S. Corrado, Jeffrey Dean. arXiv, 2013.
• [4] Zero-Shot Learning via Semantic Similarity Embedding. Ziming Zhang, Venkatesh Saligrama. ICCV, 2015.
• [5] DeViSE: A Deep Visual-Semantic Embedding Model. Andrea Frome*, Greg S. Corrado*, Jonathon Shlens*, Samy BengioJeffrey Dean, Marc’Aurelio Ranzato, Tomas Mikolov. NIPS, 2013.
• [6] Evaluation of Output Embeddings for Fine-Grained Image Classification. Zeynep Akata, Scott Reed, Daniel Walter, Honglak Lee, Bernt Schiele. CVPR, 2015.
• [7] Latent Embeddings for Zero-shot Classification. Yongqin Xian, Zeynep Akata, Gaurav Sharma, Quynh Nguyen, Matthias Hein, Bernt Schiele CVPR, 2016.
• [8] An embarrassingly simple approach to zero-shot learning. Bernardino Romera-Paredes, Philip H. S. Torr. ICML, 2015.
• [9] Label-Embedding for Image Classification. Zeynep Akata, Florent Perronnin, Zaid Harchaoui, Cordelia Schmid. PAMI, 2016.
• [10] Synthesized Classifiers for Zero-Shot Learning. Soravit Changpinyo, Wei-Lun Chao, Boqing Gong, Fei Sha. CVPR, 2016.
• [11] Semantic Autoencoder for Zero-Shot Learning. Elyor Kodirov, Tao Xiang, Shaogang Gong. CVPR, 2017.
• [12] Learning a Deep Embedding Model for Zero-Shot Learning. Li Zhang, Tao Xiang, Shaogang Gong CVPR, 2017.