To run the demo:
# Wikipedia dataset
python train_adv_crossmodal_simple_wiki.py # with contrastive loss
python train_adv_crossmodal_triplet_wiki.py # with triplet loss
We have prepared the data in the "./data" folder. If you have any question, please contact Bokun Wang at csbkwang@gmail.com.
Note: The codes were modified from the implementation of "Unsupervised Cross-modal Retrieval through Adversarial Learning", written by Li He. If you use the codes, please cite the following two papers:
[1] Li He, Xing Xu, Huimin Lu, Yang Yang, Fumin Shen and Heng Tao Shen. "Unsupervised Cross-modal Retrieval through Adversarial Learning". IEEE International Conference on Multimedia and Expo (ICME), 2017.
[2] Bokun Wang, Yang Yang, Xing Xu, Alan Hanjalic, and Heng Tao Shen. "Adversarial Cross-Modal Retrieval". In Proceedings of 25th ACM International Conference on Multimedia (ACM MM), 2017.