Object Detection
ultralytics
English
agriculture
weed-detection
yolo
yolov8
yolo26
precision-agriculture
research
Instructions to use Rmano-Rudeson/weedscan-kz with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- ultralytics
How to use Rmano-Rudeson/weedscan-kz with ultralytics:
from huggingface_hub import hf_hub_download from ultralytics import YOLO # pick the weights file from this repo's "Files and versions" tab weights = hf_hub_download("Rmano-Rudeson/weedscan-kz", "<weights>.pt") model = YOLO(weights) source = 'http://images.cocodataset.org/val2017/000000039769.jpg' model.predict(source=source, save=True) - Notebooks
- Google Colab
- Kaggle
|
Download docs/validation.md from Rmano-Rudeson/weedscan-kz: direct link, hf CLI and curl.
- Browser
- Download file 1.25 kB
-
https://huggingface.co/Rmano-Rudeson/weedscan-kz/resolve/main/docs/validation.md
- Command line
-
hf download hf://Rmano-Rudeson/weedscan-kz/docs/validation.md
-
curl -L -o validation.md https://huggingface.co/Rmano-Rudeson/weedscan-kz/resolve/main/docs/validation.md
1.25 kB
Release verification
Checked on 10 November 2026 on macOS/Apple Silicon with Python 3.13 and Ultralytics 8.4.9.
- All three released trained checkpoints loaded and completed CPU inference on the recovered frontend sample image. Label counts were 5, 15, and 10; parameter counts and detection totals are recorded in
experiments/inference-smoke-check.json. This is a functional smoke check, not an accuracy evaluation. - The FastAPI
/classifyendpoint completed real-image inference using the 15-class checkpoint, returned normalized boxes, and rejected malformed image content with HTTP 400. Weed count excludes detections whose inherited label starts withcrop:. - The recovered React client completed its production build with the locked npm dependencies. The build reports a large bundle warning and stale Browserslist metadata.
- Environment files, unused hosted-service scaffold, build output, local environments, and installed dependencies are excluded from the source release.
- Original CSV logs are retained, and SHA-256 checksums identify the uploaded checkpoint bytes.
No independent test-set accuracy, field deployment, treatment recommendation, or dataset-license audit was performed. Raw datasets are not part of this release.