---
license: agpl-3.0
library_name: ultralytics
pipeline_tag: image-segmentation
tags: [yolo, yolo26, onnx, instance-segmentation, coco]
---

# yolo26n-seg (ONNX)

The official Ultralytics **YOLO26n-seg** COCO weights (`yolo26n-seg.pt`), exported unchanged to ONNX so they can run in the
browser with onnxruntime-web. Not fine-tuned.

| | |
|---|---|
| Input | `images` — float32 `[1, 3, 640, 640]`, RGB 0–1, letterboxed (pad 114) |
| Output 0 | `[1, 116, 8400]` = box `cx, cy, w, h` + 80 class scores + 32 mask coefficients. NMS is **not** included |
| Output 1 | `[1, 32, 160, 160]` mask prototypes — mask = sigmoid(coefficients · prototypes), cropped to the box |
| Size | ONNX 11.2 MB (opset 17, simplified) · source `.pt` 6.7 MB |
| Classes | COCO 80 |

Used by the model-card site [AI Models Organize](https://ai-models-organize.vercel.app/) (YOLO26n-seg card → «돌려보기»).

## License

AGPL-3.0, inherited from Ultralytics YOLO.