Optimized Pedestrian Detection Leveraging YOLOv9: A Thorough Deep Learning Framework for Autonomous Vehicle System
Hemant Kumar, Pushpa Mamoria, Deepak Kumar Dewangan · 2024
Pedestrian detection for an autonomous vehicle system is the crucial and foremost important aspect for accurate identification of pedestrian on the road scene environment. For this study a YOLOv9-C state-of-the-art deep learning model is used. YOLOv9 is renowned for its accuracy and efficiency in real time application. Model is trained on KITTI dataset. This dataset has a number of classes but for this study only pedestrian class is used. Dataset provides a number of images of urban scene environment where number of objects are there and trained model is perfectly detecting pedestrian in the images. In the YOLOv9 model an additional backbone layer is added for achieving high efficiency and time bound detection. The primary aim of this research is to look for the possible technique by which detection accuracy can be improved. Additional added layer helps in detecting more complex features of pedestrian in the images. After careful observation findings are shows that proposed model has achieved better accuracy in comparison with another models. Experimental results also show that there is a chance for more reliability and responsiveness of object detection if model is trained on diverse dataset. Although this study demands high computational resources which may be a barrier in some sense.