Research on the Application of CIoU and CNN Joint Compression in Autonomous Driving

With the rapid development and application of autonomous driving technology, vehicle computing devices are facing increasingly daunting challenges. These systems need to process a large amount of sensory data in real-time, plan routes, and make decisions, which requires more computational resources. Therefore, effective model compression methods have become crucial. Our study aims to reduce the size and computational complexity of models through weight pruning and convolutional split compression methods, thereby reducing the consumption of computational resources while maintaining model performance and accuracy. We also utilize the CIoU loss function to improve the accuracy and efficiency of obstacle detection in autonomous driving systems. In simulated experiments, we tested and evaluated the improved model using a dataset of simulated road scenarios. The results of the experiments validate the effectiveness of model compression and demonstrate the advantages of the improved model in increasing obstacle detection rates and reducing computational costs. These improvements can help autonomous driving systems operate efficiently with limited computational resources, thereby enhancing system reliability and practicality. Our research provides valuable theoretical and practical support for advancing autonomous driving technology.