In-house model and microservice for detecting varroa mites directly on bees in hive images.
Repository: https://github.com/Gratheon/models-varroa-on-bee
It is integrated in our pipeline (web-app -> graphql-router -> image-splitter -> models-varroa-on-bee) and returns bounding boxes over HTTP.
Highlights:
- Dedicated
varroa_on_beedetections (not only hive-bottom mites) - Simple API (
POST /withmultipart/form-dataimage upload) - Health endpoint for operations (
GET /health)
Validation metrics (varroa_model5, best.pt):
- Precision:
0.926 - Recall:
0.823 - mAP50:
0.871 - mAP50-95:
0.485 - Varroa class precision/recall:
0.858/0.651 - Dataset source: Roboflow Universe
varroa-j8231/varroa8kv1
https://universe.roboflow.com/varroa-j8231/varroa8k/dataset/1