Human eye-fixation prediction based on Convolutional Neural Network in RGB images
Miao yue, Wei Li, LI Chunle · Journal of Physics Conference Series · 2021
Abstract Using neural networks to simulate and predict human visual attention mechanism is a hot topic in the field of computer vision. In this paper, we propose an end-to-end encoder-decoder network architecture to predict the fixation mechanism of human eyes, which consists of modules composed of multiple convolutional layers with different expansion rates to capture multi-scale features in parallel. In addition, the attention module is added on the basis of the encoder network structure, and a self-attention mechanism is introduced to capture the visual feature dependency in the channel size, and the semantic interdependence in the channel dimension is modeled to predict the visual salience more accurately. In this paper, five data sets and selected examples are used to demonstrate the effectiveness of the proposed method. Our method achieves competitive and consistent results on multiple evaluation indicators on MIT1003 and CAT2000 datasets.