The Japanese Society for Artificial Intelligence

This study aims to achieve better model results with a reduced volume of data through data fusion methods. By integrating road image data and eye movement data, we have constructed a custom dataset tailored for the traffic domain. This dataset will be applied to the YOLO v5 and v8 models, creating a multifunctional and universal data fusion architecture, aimed at reducing the data requirements of current object detection models. This method is intended to improve the final accuracy and processing speed of the models, particularly in real-time applications where computational resources and time are limited. By optimizing the data processing pipeline, we ensure that the fused data is not only more compact but also more relevant and informative for object detection tasks. This data fusion strategy offers a scalable solution that balances data efficiency with high accuracy. This study’s objective is to validate the effectiveness of the proposed data fusion method in enhancing the performance of object detection systems, which stands as a significant contribution to the computational efficiency in the data science and computer vision domains.