Target Search via Feature Cutting Strategy of Visual Attention Mechanism
Fang Guo, Jianyu Zhao, Ping Jiang · 2019
Only part of the information and specific goals need to be paid attention to in the real image processing. Based on this a method using top-down visual attention mechanism is proposed in this paper to find a significant part and the given target with the learning of different features in different scales of sample images. The weights of the extracted features in different scales are calculated. Only the important features and optimal scales are selected in the real processing, that is the features and scales are cut when the relative weights are lower than the given threshold. The experimental results show that the presented method can localize the target accurately and quickly.