PLDH: Pseudo-Labels Based Deep Hashing

Deep hashing has received a great deal of attraction in large-scale data analysis, due to its high efficiency and effectiveness. The performance deep models heavily relies on label information, which is very expensive obtain. In this work, novel end-to-end model based pseudo-labels for without labels proposed. proposed consists two major stages, where the first stage aims obtain features extracted by pre-training convolution neural network. second generates hash codes with quality same network previous stage, coupled an layer, whose purpose encode into binary representation. Additionally, quantization loss introduced interwound within these stages. Evaluation experiments were conducted frequently-used image collections, CIFAR-10 NUS-WIDE, eight popular shallow models. experimental results show superiority method retrieval.