Designing an Efficient Object Detection Model for Autonomous Driving Applications

The detection of road objects plays an essential role in the development of autonomous vehicles in intelligent transportation systems, which has become an emerging field in deep learning. Several objects on the road, like vehicles, pedestrians, etc., are necessary to be accurately identified, which guarantees the safety of other people and vehicles in the surroundings. In this article, we aim to design an efficient object detection model for autonomous driving systems. To achieve this, we investigate the recently developed YOLOv7 and optimize the same for improving the detection performance to satisfy the realtime safety requirements of autonomous vehicles. We perform extensive experimentation and demonstrate the effectiveness of our method on the BDDIOOK dataset. Experimental results show that our proposed method increases the detection accuracy to 82.6% and inference speed to 97.2 FPS compared to the baselines, with no additional increase in model complexity.