Driver attention prediction based on convolution and transformers

In recent years, studying how drivers allocate their attention while driving is critical in achieving human-like cognitive ability for autonomous vehicles. And it has been an active topic in the community of human-machine augmented intelligence for self-driving. However, existing state-of-the-art methods for driver attention prediction are mainly built upon convolutional neural network (CNN) with local receptive field which has a limitation to capture the long-range dependencies. In this work, we propose a novel Attention prediction method based on CNN and Transformer which is termed as ACT-Net. In particular, CNN and Transformer are combined as a block which is further stacked to form the deep model. Through this design, both local and long-range dependencies are captured that both are crucial for driver attention prediction. Exhaustive comparison experiments over other state-of-the-art techniques conducted on widely used dataset of BDD-A and private collected data on BDD-X validate the effectiveness of the proposed ACT-Net.