This paper addresses the detection requirements for rear floor components in Chery Automobile’s welding workshop by proposing an enhanced detection algorithm based on YOLOv11. This algorithm integrates advanced attention mechanisms and feature fusion techniques. To tackle the challenge of accurately detecting densely distributed multiple targets in complex industrial environments, we introduce three improved variants: (1) C2f-SimAM variant, which enhances feature extraction by integrating a spatial-aware attention module; (2) C3k2-CMUNeXt variant, which improves feature representation using large-kernel deep convolutions; (3) a comprehensive model that synergistically integrates both enhancements. Comprehensive experiments on a dataset containing 25 industrial parts demonstrate that our variants significantly outperform the baseline YOLO11 model. Under stringent industrial constraints ( ≤50M parameters, ≥50FPS inference speed), the C3k2-CMUNeXt variant with the best performance achieved 92.7 % mAP @ 50 and 72.0 % mAP @ 0.5 : 0.95, which was 7.0 % higher than the baseline. Comparative analysis with six classic detection models validates the superiority of our approach. The proposed method achieves state-of-the-art accuracy for industrial underfloor component detection while satisfying real-time performance and model size constraints.
Tianyi Lai (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: