Bias: Bijective Input And Surjectivity In Zero Shot Learning

Zero-shot learning suffers from the issue of generalisation due to domain shift across seen and unseen classes. In this paper, we propose a method that extends the usual approach of learning a mapping between semantic and visual embedding spaces by ensuring it to be surjective. This functional constraint along with triplet loss prevents the model from overfitting to seen classes. We also use a bijective feature extractor to complement our proposal. Experimental results on benchmark datasets depict that our method out-performs standard approaches in conventional and generalised scenarios.