NCTU

Deep learning of feature representations for efficient retrieval

2017/09/13
CodeX17-CSC-00-1
Professorprofessor Chu-Song Chen
ApplicationEntertainment, Medicine and health care, Surveillance and security, the applications of education, and Vehicle
Technical BenefitImprove efficiency/ increasing execution speed
Technology StatusCan be transferred

To improve the retrieval efficiency of deep networks, we proposed a simple yet effective supervised deep hash approach that constructs binary hash codes from labeled data for large-scale image search. Our approach constructs hash functions as a latent layer in a deep network. The learned features can help increase the retrieval speed.

Figure: SSDH takes inputs from images and learns image representations, binary codes, and classification through the optimization of an objective function that combines a classification loss with desirable properties of hash codes. The learned codes preserve the semantic similarity between images and are compact for image search.