DIFAIR: Towards learning DIFerentiAted Image Representations

DIFAIR (DIFferentiAted Image Representations) is an approach to learn a specific representation for deep neural networks applied to image classification. The objectives are to obtain representations exhibiting: (i) class separability, through predefined class positions in the representation space; (ii) the extraction of distinct features, which remain inactive if not present in the image; and (iii) semantic meaning when comparing representations. A distance-based loss function is proposed to optimize a network, in a supervised way, to obtain the desired representation. 

This resource contains additional figures containing examples of representations for different images.