pub fn decode_yolo_split_end_to_end_det_float<T: Float + AsPrimitive<f32>>(
boxes: ArrayView2<'_, T>,
scores: ArrayView2<'_, T>,
classes: ArrayView2<'_, T>,
score_threshold: f32,
output_boxes: &mut Vec<DetectBox>,
) -> Result<(), DecoderError>Expand description
Decodes split end-to-end YOLO detection outputs (post-NMS from model).
Input shapes (after batch dim removed):
- boxes: (N, 4) — xyxy pixel coordinates
- scores: (N, 1) — confidence of the top class
- classes: (N, 1) — class index of the top class
Boxes are output directly without NMS (model already applied NMS).