Attribute Recognition from Adaptive Parts
This work addresses attribute recognition for computer vision applications, presenting an incremental improvement over previous part-based methods.
The paper tackles the problem of sub-optimal part detection in attribute recognition by proposing an end-to-end deep learning approach that jointly learns key point estimation and attribute recognition, achieving verified efficacy on two datasets.
Previous part-based attribute recognition approaches perform part detection and attribute recognition in separate steps. The parts are not optimized for attribute recognition and therefore could be sub-optimal. We present an end-to-end deep learning approach to overcome the limitation. It generates object parts from key points and perform attribute recognition accordingly, allowing adaptive spatial transform of the parts. Both key point estimation and attribute recognition are learnt jointly in a multi-task setting. Extensive experiments on two datasets verify the efficacy of proposed end-to-end approach.