Efficient Optimization and Compression of Autonomous Vehicle Target Detection Models

Target detection is a crucial element of computer vision and often requires complex models and extensive computing resources. This inherent complexity poses significant challenges for use especially in the demanding field of autonomous vehicles. Thus, for the urgent need to reduce resource consumption while preserving the accuracy of target detection, we optimized the YOLO model by utilizing depthwise separable convolution. The model reduced resource usage without compromising detection performance. Additionally, quantitative compression techniques were incorporated to decrease the model size for easy deployment. The optimization strategy better-tailored target detection solutions to the specific requirements of autonomous vehicle applications.