Skip to main content
. 2025 Sep 12;8:1637842. doi: 10.3389/frai.2025.1637842

Figure 2.

Block diagram of the YOLOv10 network, divided into backbone, neck, and head. The backbone stack shows layers such as Conv, C2f, SCDown, C2f/CIB, SPPF, and PSA arranged from shallow to deep. The neck connects multi-scale features using Concat and Upsample operators. The head splits into two branches labeled “One-to-one Head” and “One-to-many Head,” indicating different assignment strategies. Arrows convey the feature flow across scales. Modules are grouped by region, with consistent labels and panel boundaries to emphasize the three-part structure.

YOLOv10.