Accelerating Image Analysis Using AI-Driven Methods: Enhancing Speed and Accuracy in Autonomous Vehicle System

الملخص

The fast improvements in Autonomous Vehicle (AV) systems have shown the necessity for effective image processing approaches to enable efficient decision-making. Current approaches generally fail to offer a balance between processing speed and precision, restricting their usefulness in AV scenarios. The work sets out to tackle these difficulties by applying advanced AI-driven methodologies, EfficientNet and MobileNet, to optimize picture analysis for AV systems. This research filled a breach by enhancing both the speed and accuracy of real-time image processing systems. Also, it contributed to scientific studies in this sector. According to the KITTI Vision Benchmark Suite and Berkeley DeepDrive datasets, the experimental quantitative research designs. Proposed models were trained and tested using these datasets. TensorFlow and Keras frameworks incorporated advanced convolutional neural network topologies with transfer learning algorithms. The models were released loose under varied driving circumstances to see how flexible and resilient they were. The statistical importance of performance parameters like accuracy, inference time, and F1-score was evaluated. The results reveal that EfficientNet can obtain an accuracy of 94.2% and an inference time of 18 ms/image, which is substantially better than the baseline. MobileNet was a plausible option, exhibiting amazing accuracy while being computationally efficient. This improvement was statistically significant, and qualitative assessments indicated that the models were powerful under bad conditions. The research advances real-time imaging analysis in AVs, pointing to the need for architectural adjustments and dataset diversity. As a result of this research, the field of AI-controlled image processing will advance and lead to creative developments in AV systems and their applications.