Object detection is a crucial component of the perception system in autonomous driving. However, the road scene presents a highly intricate environment where the visibility and characteristics of traffic targets are susceptible to attenuation and loss due to various complex road scenarios such as lighting conditions, weather conditions, time of day, background elements, and traffic density. Nevertheless, the current object detection network must exhibit more learning capabilities when detecting such targets. This also exacerbates the loss of features during the feature extraction and fusion process, significantly compromising the network’s detection performance on traffic targets. This paper presents a novel methodology by which to overcome the concerns above, namely HRYNet. Firstly, a dual fusion gradual pyramid structure (DFGPN) is introduced, which employs a two-stage gradient fusion strategy to enhance the generation of more comprehensive multi-scale high-level semantic information, strengthen the interconnection between non-adjacent feature layers, and reduce the information gap that exists between them. HRYNet introduces an anti-interference feature extraction module, the residual multi-head self-attention mechanism (RMA). RMA enhances the target information by implementing a characteristic channel weighting policy, thereby reducing background interference and improving the attention capability of the network. Finally, the detection performance of HRYNet was evaluated by utilizing three datasets: the horizontally collected dataset BDD1000K, the UAV high-altitude dataset Visdrone, and a custom dataset. Experimental results demonstrate that HRYNet achieves a higher mAP_0.5 compared with YOLOv8s on the three datasets, with increases of 10.8%, 16.7%, and 5.5%, respectively. To optimize HRYNet for mobile devices, this study presents Lightweight HRYNet (LHRYNet), which effectively reduces the number of model parameters by 2 million. The results demonstrate that LHRYNet outperforms YOLOv8s in terms of mAP_0.5, with improvements of 6.7%, 10.9%, and 2.5% observed on the three datasets, respectively.

Although the vision transformer has been used in gait recognition, its application in multi-view gait recognition remains limited. Different views significantly affect the accuracy with which the characteristics of gait contour are extracted and identified. To address this issue, this paper proposes a Siamese mobile vision transformer (SMViT). This model not only focuses on the local characteristics of the human gait space, but also considers the characteristics of long-distance attention associations, which can extract multi-dimensional step status characteristics. In addition, it describes how different perspectives affect the gait characteristics and generates reliable features of perspective–relationship factors. The average recognition rate of SMViT for the CASIA B dataset reached 96.4%. The experimental results show that SMViT can attain a state-of-the-art performance when compared to advanced step-recognition models, such as GaitGAN, Multi_view GAN and Posegait.

Aiming at challenges such as the high complexity of the network model, the large number of parameters, and the slow speed of training and testing in cross-view gait recognition, this paper proposes a solution: Multi-teacher Joint Knowledge Distillation (MJKD). The algorithm employs multiple complex teacher models to train gait images from a single view, extracting inter-class relationships that are then weighted and integrated into the set of inter-class relationships. These relationships guide the training of a lightweight student model, improving its gait feature extraction capability and recognition accuracy. To validate the effectiveness of the proposed Multi-teacher Joint Knowledge Distillation (MJKD), the paper performs experiments on the CASIA_B dataset using the ResNet network as the benchmark. The experimental results show that the student model trained by Multi-teacher Joint Knowledge Distillation (MJKD) achieves 98.24% recognition accuracy while significantly reducing the number of parameters and computational cost.

Color fusion and rivalry are two key information integration mechanisms in binocular vision, representing the visual system’s processing patterns for consistent and conflicting inputs, respectively. This study hypothesizes that there are quantifiable differences in eye movement indicators under states of binocular color fusion and rivalry, which can be verified through multi-paradigm eye movement experiments. The experiment recruited eighteen subjects with normal vision (nine males and nine females), employing the Gaze Stability paradigm, Straight Curve Eye Hopping paradigm, and Smoothed Eye Movement Tracking paradigm for eye movement tracking. Each paradigm included a binocular color rivalry experimental group (R-G) and two binocular color fusion control groups (R-R, G-G). Data analysis indicates significant differences in indicators such as Average Saccade Amplitude, Median Saccade Amplitude, and SD of Saccade Amplitude between binocular color fusion and rivalry states. For instance, through Z-Score normalization and cross-paradigm merged analysis, specific ranges of these indicators were identified to distinguish between the two states. When the Average Saccade Amplitude falls within the range of −0.905–−0.693, it indicates a state of binocular color rivalry; when the range is 0.608–1.294, it reflects a state of binocular color fusion. Subsequently, ROC curve analysis confirmed the effectiveness of the experimental paradigms in analyzing the mechanisms of binocular color fusion and rivalry, with AUC values of 0.990, 0.741, and 0.967, respectively. These results reveal the potential of eye movement behaviors as biomarkers for the dynamic processing of visual conflicts. This finding provides empirical support for understanding the neural computational models of binocular vision and lays a methodological foundation for developing visual impairment assessment tools based on eye movement features.