Detection of Forklift Trucks in Warehouses using Transfer Learning Techniques

The Industrial Revolution introduced mechanized production, leading to significant advancements in technology, economy, society, and culture. This shift necessitated the development of efficient storage and distribution systems, positioning warehouses as central nodes in the supply chain. Forklifts, a pivotal tool in modern warehousing, revolutionized material handling by enabling the efficient movement and stacking of heavy loads. However, detecting forklifts in busy and confined warehouse environments is crucial for ensuring safety, optimizing operations, and preventing accidents. This study evaluates the performance of various You Look Only Once (YOLO) models (v5, v8, v9, and v10) in detecting forklifts in warehouses, comparing their training and validation losses, precision, recall, and mean Average Precision (mAP). YOLO v8 demonstrated superior stability and consistency across these metrics, making it the preferred model for forklift detection. While YOLO v9 and v10 exhibited potential for handling more complex scenarios, they showed greater fluctuations in performance, indicating the need for careful monitoring and tuning. The findings suggest that YOLO v8 is the most reliable choice for enhancing safety and operational efficiency in warehouse environments, with YOLO v9 and v10 offering potential benefits in specific, more complex applications.