TARGET OBJECT DETECTION METHOD, APPARATUS AND SYSTEM AND NEURAL NETWORK STRUCTURE

摘要

A target object detection method, apparatus and system, and a neural network structure, the method comprising: determining at least one region of interest to be detected at a current frame of a video image sequence (S100), each region of interest at least partially comprising information of at least one target object; respectively extracting features of a target object in the at least one region of interest of the current frame (S200); predicting at least one region of interest of the current frame according to the features of the target object so as to obtain a prediction result (S300); and determining a region of interest to be detected of a subsequent frame according to the prediction result (S400). Thus, when a target object is detected, the information of a current frame may be transmitted to a subsequent frame so as to reuse information of a time domain between different images, and long-range time-domain features are fully utilized so as to provide a time-domain basis for complicated situations such as processing appearance changes of an object.