Deep Learning Model To Detect Driver Hand Gestures And Vehicle Signals in Indian Traffic

Abstract: This research aims to improve road safety by developing a deep learning-based system capable of detecting and recognizing tail lights, brake lights, and driver hand gestures in Indian traffic conditions. Using the YOLOv8 object detection architecture, the system has been designed and implemented in two key phases: training and inference. During the training phase, a comprehensive custom dataset of traffic images - carefully labelled with seven distinct classes - was used to train the YOLOv8 model. This dataset includes real-world conditions such as varying lighting, complex backgrounds, and occlusions. In the inference phase, the trained model processes new images or videos, automatically detecting the presence of vehicle signals and hand gestures. The output consists of bounding boxes and class labels that are visually rendered and saved, ensuring traceability and ease of analysis.

Keywords: Brake Light Detection, Tail Light Detection, Hand Gesture Recognition, Object Detection.