Sketch-Based Image Retrieval via Compact Binary Codes Learning

With the exploding number of images on the Internet and the convenience of free-hand sketch drawing, sketch-based image retrieval (SBIR) has attracted much attention in recent years. Due to the ambiguity and sparsity of sketches, SBIR is more challenging to cope with than conventional content-based problem. Existing approaches usually adopt high-dimensional features which require high-computational cost. Furthermore, they often use edge detection and parameter-sharing networks which may lose important information in training. In this study, we propose a compact binary codes learning strategy using deep architecture. By leveraging well-designed prototype hash codes, we embed different domains input (sketch and photo) into a common comparable feature space. Besides, we present two separate networks specific to sketches and real photos which can learn very compact features in Hamming space. Our method achieves state-of-the-art results in accuracy, retrieval time and memory cost on two standard large-scale datasets.