Eye Fixation Assisted Video Saliency Detection via Total Variation-based Pairwise Interaction

A Study of Eye-Tracking Gaze Point Classification and Application Based on Conditional Random Field

The head-mounted eye-tracking technology is often used to manipulate the motion of servo platform in remote tasks, so as to achieve visual aiming of servo platform, which is a highly integrated human-computer interaction effect. However, it is difficult to achieve accurate manipulation for the uncertain meanings of gaze points in eye-tracking. To solve this problem, a method of classifying gaze points based on a conditional random field is proposed. It first describes the features of gaze points and gaze images, according to the eye visual characteristic. An LSTM model is then introduced to merge these two features. Afterwards, the merge features are learned by CRF model to obtain the classified gaze points. Finally, the meaning of gaze point is classified for target, in order to accurately manipulate the servo platform. The experimental results show that the proposed method can classify more accurate target gaze points for 100 images, the average evaluation values Precision = 86.81%, Recall = 86.79%, We = 86.79%, these are better than relevant methods. In addition, the isolated gaze points can be eliminated, and the meanings of gaze points can be classified to achieve the accuracy of servo platform visual aiming.

Bio-Inspired Representation Learning for Visual Attention Prediction

Visual attention prediction (VAP) is a significant and imperative issue in the field of computer vision. Most of the existing VAP methods are based on deep learning. However, they do not fully take advantage of the low-level contrast features while generating the visual attention map. In this article, a novel VAP method is proposed to generate the visual attention map via bio-inspired representation learning. The bio-inspired representation learning combines both low-level contrast and high-level semantic features simultaneously, which are developed by the fact that the human eye is sensitive to the patches with high contrast and objects with high semantics. The proposed method is composed of three main steps: 1) feature extraction; 2) bio-inspired representation learning; and 3) visual attention map generation. First, the high-level semantic feature is extracted from the refined VGG16, while the low-level contrast feature is extracted by the proposed contrast feature extraction block in a deep network. Second, during bio-inspired representation learning, both the extracted low-level contrast and high-level semantic features are combined by the designed densely connected block, which is proposed to concatenate various features scale by scale. Finally, the weighted-fusion layer is exploited to generate the ultimate visual attention map based on the obtained representations after bio-inspired representation learning. Extensive experiments are performed to demonstrate the effectiveness of the proposed method.