Attention Transition Prediction Based on Multi-Scale Spatiotemporal Features

The human visual system has powerful information processing capabilities, and humans can find their areas of interest in a very short time. The traditional salient object detection method is dedicated to detecting the area where the salient object is located. But when people are concentrating, they often only focus on a part of an object, not the whole object. Therefore, detecting the gaze position by detecting the eye movement can better simulate the attention transition. In order to achieve a more accurate simulation of human visual attention mechanism, we propose a model to predict attention transition of consecutive frames of video, which use multi-scale spatiotemporal features to extract static and dynamic saliency maps separately.