Multi-dimensional Equilibrium-depth Hashing Method for Large-Scale Image Retrieval

Aiming at the limited feature extraction capability and inefficient quantization constraint mechanism of existing hashing methods,a deep multi-scale attention hashing network was proposed for large-scale image retrieval . The equilibrium network consists of two sub-networks, the main branch and the object branch, and adds a multi-dimensional significant area extraction module to the main branch network to effectively extract the saliency region in the image, and send the execution results to the object branch network to learn more detailed features. Triplet quantization constraints are introduced to reduce the quantization error and preserve the similarity relationship of pairs of samples. To verify the effectiveness of the method, extensive experiments are performed on two benchmark datasets. Experimental results show that the proposed method outperforms most existing hash retrieval methods.