README.md

---
license: agpl-3.0
library_name: ultralytics
pipeline_tag: object-detection
tags:
  - darts
  - dartboard
  - yolo11
  - onnx
datasets:
  - geforcefan/dartscribe
---

# dartscribe networks

The three networks of [dartscribe](https://github.com/geforcefan/dartscribe). Two calibrate: they find the bull and
the crossings of a dartboard in an image and tell where the 20 is. One finds the dart tips.

| file | network | input | output |
|---|---|---|---|
| `crossings.onnx` | YOLO11n detect | 1×3×960×960 | 1×9×18900 |
| `orientation.onnx` | YOLO11n classify | 1×3×512×512 | 1×20 |
| `darts.onnx` | YOLO11n detect | 1×3×1280×1280 | 1×5×33600 |

RGB, values divided by 255, NCHW, no NMS in the ONNX.

## Crossings

Input: the image letterboxed to 960×960.

Output: one column per candidate box.

| row | value |
|---|---|
| 0, 1 | center x, y |
| 2, 3 | width, height |
| 4 | bull |
| 5 | double outer |
| 6 | double inner |
| 7 | treble outer |
| 8 | treble inner |

The classes do not name the crossing, the board is symmetric.

## Orientation

Input: the board warped into a 512×512 square with the projection, board center in the middle, 1.6 double outer radii
to the edge.

Output: one score per class, class k means the 20 is k segments clockwise from the top.

## Darts

Input: the image letterboxed to 1280×1280.

Output: one column per candidate box.

| row | value |
|---|---|
| 0, 1 | center x, y |
| 2, 3 | width, height |
| 4 | tip |

## Letterbox

The image is scaled into the square with its aspect ratio kept, the rest is gray (114), the YOLO convention.