Table 2 Comparison of YOLOv5 and Hybrid-YOLOv5 with architectural modifications and motivations.
From: Hybrid-YOLOv5 for object detection of non-ferrous metals in end-of-life vehicles
Component | YOLOv5 (Original) | Hybrid-YOLOv5 (Proposed) | Motivation for Change |
|---|---|---|---|
Backbone | CSPDarknet53 | MobileNetV3 + CSPDarknet53 | Improve computational efficiency while maintaining sufficient detection accuracy. |
Backbone Module | C3 (Cross Stage Partial Module) | C2 F (Coarse-to-Fine Module) | Enhance hierarchical feature extraction, improving small object detection and texture representation. |
Attention Mechanism | None | SE (Squeeze-and-Excitation) Module | Improve feature recalibration and emphasize critical features in complex visual tasks. |
