Scanpaths Prediction Based on Signals Competition

Human scanpath is an important reflection of the cognitive process in image observation and understanding. Accurately predicting scanpaths that match the physiological mechanism of the human eye under visual search tasks is helpful for designing a more natural human-computer interaction system. Aiming at the problem that current physiologically inspired scanpaths prediction model has simple input information and does not consider the interaction and competition of different signals, this paper proposes a scanpaths prediction model based on the signals competition. In this model, top-down and bottom-up signals are obtained from a two-stage object detection network, and a "parallel processing - organic fusion" computing structure is established based on physiological parameters to simulate the visual search process. In this paper, the model is tested on the dataset of clock and microwave(MCS), and the results show that the method has a significant improvement over the benchmark in the metrics.