BullTrack v0.1.0

First public release.

BullTrack is an Android app that scores steel-tip and soft-tip darts with the phone camera: mount the phone beside the board and each dart is detected and scored as it lands.

Highlights

• Auto scoring with drag-to-correct on a virtual board
• Placement guide and automatic zoom to the board
• X01, the Cricket family, Shanghai, Half It, Bermuda, Gotcha 301, Killer, Catch 40, 41-60 and practice games; 1-4 players or 2-4 teams
• Per-player stats, head-to-head records, hit maps and share images
• Sound cues for doubles, trebles and the bull
• Match history backup and restore; English and Chinese UI

Requirements

• 64-bit Android 8.0 or later with a rear camera
• Detection runs on the GPU (Vulkan) when it passes a self-check against the CPU, otherwise on the CPU

Files

File	SHA-256
BullTrack-v0.1.0-android-arm64-v8a.apk	7b9b89c5db6ed8531cbf2042096df11f27588ab9ead450b75b67fa247c5b2103

The APK was built and signed by GitHub Actions from this tag. To check that it came from this repository:

gh attestation verify BullTrack-v0.1.0-android-arm64-v8a.apk --repo Thomasff/BullTrack

The APK already contains the detection model. It uses the model from model-v1, which you need only to build from source.

Known limitations

• Very transparent or pale shafts and flights, or boards with unusual colours, may not be detected.
• Darts landing very close together or hidden behind each other are occasionally misread and need correcting.
• Tested on only a few phones so far. If you run into problems on yours, please report them.

If detection goes wrong, please open an issue with a debug bundle (see "Reporting issues" in the README).

---

首个公开版本。用手机摄像头给钢镖、软镖自动计分，支持中英文界面。说明见 中文 README。

model-v1

Detection model for building BullTrack from source. App releases already contain it; you don't need these files to install the app.

Used by: v0.1.0 and later, until a newer model release says otherwise.

File	SHA-256
model.ncnn.param	454a5a2c040c966149dc7785d30731a213b7ff381812509ea5f46e1f7c87de8c
model.ncnn.bin	bc04f7cc77d19471d8885ec350335819f250f44fb05977c91dc2dbbd20ebc913

Put both files in bulltrack-core/src/main/assets/ncnn/ before building (see "Building from source" in the README).

Fine-tuned from the Ultralytics YOLO11s pretrained weights and converted to ncnn. Distributed under AGPL-3.0, see THIRD_PARTY_NOTICES.md.