Industrial-YOLO detects defects at 120 FPS on edge hardware

Automated surface defect detection is critical for quality control in high-speed manufacturing, but deploying deep learning models on resource-constrained edge hardware typically introduces latency. A new paper by Emmanuel Ezeji Somtochukwu and Nitesh Rijal tackles this with Industrial-YOLO, an edge-optimized framework built on a fine-tuned YOLOv8 architecture. The framework incorporates target-specific optimizations via TensorRT and OpenVINO acceleration engines to bridge the gap between algorithmic complexity and hardware constraints.

In systematic benchmarks using the NEU surface defect database (steel sheets) and MVTec AD dataset, supplemented with custom automotive manufacturing extensions (scratches, pits, inclusions), Industrial-YOLO achieved a high-velocity inference speed exceeding 120 FPS on the NVIDIA Jetson Orin platform while maintaining an exceptional mean Average Precision (mAP) of 98.5%. The system demonstrated robust, zero-latency performance when deployed directly onto an active automotive assembly line, offering a scalable blueprint for next-generation automated optical inspection systems. This work provides a practical path to real-time, on-device quality control without cloud dependency.