Shoeprint Image Retrieval Based on Dual Attention Light Hash Network
Daxiang Li, Yang Li, Ying Liu · 2021 4th International Conference on Artificial Intelligence and Pattern Recognition · 2021
Aiming at the problem of the limited computing power of public security Internet of Things (IoT) mobile terminals, a new dual-attention light hash network is designed to realize the rapid retrieval of crime scene Investigation shoeprint images. First, a spatial attention (SA) module is constructed using deformable convolution to pay more attention to the significant regions of the shoeprint image; then, a channel attention (CA) module is constructed using a dimensionless local cross-channel interaction strategy to perform adaptive weighting of different channels of the feature mapping. Finally, a hash module is constructed using the slicing method to transform the real-valued features into a binary output as the final representation of the shoeprint image. The generated hash codes are constrained in the siamese network to enhance their distinguishing ability. We test our method on the CSFID and FID-300 datasets and we compare our method with other state-of-the-art methods in the shoeprint image retrieval domain. The experimental results show that our method can improve the image retrieval baseline by a large margin and better than other methods.