An Efficient Supervised Deep Hashing Method for Image Retrieval

In recent years, searching and retrieving relevant images from large databases has become an emerging challenge for the researcher. Hashing methods that mapped raw data into a short binary code have attracted increasing attention Most existing hashing approaches map samples to vector via single linear projection, which restricts flexibility of those leads optimization problems. We introduce CNN-based method uses multiple nonlinear projections produce additional short-bit tackle this issue. Further, end-to-end system is accomplished using convolutional neural network. Also, we design loss function aims maintain similarity between minimize quantization error by providing uniform distribution hash bits illustrate proposed technique’s effectiveness significance. Extensive experiments conducted on various datasets demonstrate superiority in comparison with state-of-the-art deep methods.