Module yolo26 

The YOLO26 model architecture: blocks, variants, and the assembled network. YOLO26 detection model.

Architecture reference: ultralytics/cfg/models/26/yolo26.yaml.

Returns raw DetectOutput { boxes (B, 4·A), scores (B, nc·A) } — the training-mode layout. For inference, apply detect_decode on boxes using the anchor grid and strides [8, 16, 32] appropriate for the runtime input resolution.

Modules

blocks

Building blocks (backbone/neck/head layers) used to assemble YOLO26.

Structs

Yolo26Config

Per-variant scaling parameters derived from the yaml model definition.

Enums

Yolo26Variant

YOLO26 model size variant.

Functions

yolo26

YOLO26 single-head forward closure (inference or one-head training).

yolo26_dual

YOLO26 dual-head forward closure for training with consistent dual assignment.