Semi‐Paired Semi‐Supervised Deep Hashing for cross‐view retrieval

Due to its fast computational speed and low storage cost, hashing has been effectively applied large-scale multimedia retrieval tasks, such as medical video security retrieval. Most existing cross-view methods require good matching information, however, this exact pairing relationship is difficult fully realise in practice. The association between views incomplete, the label information. This task of missing paired labelled information very challenging, but less explored research. In study, a semi-supervised semi-paired deep for data proposed, named Semi-Paired Semi-Supervised Deep Hashing (SPSDH) solve challenging task. SPSDH novel end-to-end neural network model with high-order affinity. A non-local higher-order affinity measure that better considers multimodal neighbourhood structure proposed. common representation associate different modalities introduced, which combined greatly maintains consistency within modalities. evaluated on three benchmark datasets approximate nearest neighbour search compared several state-of-the-art methods. Extensive experimental results demonstrate superior performance our proposed tasks.