Smart Warehouse Safety: Computer Vision for Forklift Driver Monitoring in Warehouse Setting

Safety for forklifts and warehouse workers must be prioritized. Our study introduces a computer vision-based solution for promptly recognizing unsafe behaviors in real time using deep learning and pose estimation. The primary goal is to identify unsafe actions of forklift drivers, enhance workplace safety, and help managers spot dangerous trends or patterns early on. The research will integrate Convolutional Neural Networks and Pose Estimation. Specifically, we will use the “You Only Look Once” architecture version 8 of the CNN model for object detection to identify forklift drivers. Subsequently, the YOLO version 8 pose will detect and estimate forklift driver body joints. Our findings indicate that this approach achieves a detection accuracy rate of over 91 percent and an f1 score of the pose estimation model exceeding 0.96. Importantly, the potential of our solution is not limited to forklift safety in warehouses. Future work will focus on adapting the model to different warehouse settings and other high-risk industries, demonstrating the versatility and wide-ranging applicability of our research.