Public project · Computer vision
Hymenoptera Recognizer
Recognizing bees, wasps, and hornets in images or video.
Context
The project
Hymenoptera Recognizer trains and uses a YOLOv5 model to distinguish five categories: bee, wasp, European hornet, Asian hornet, and Oriental hornet. The repository covers images, videos, and webcam streams.
Objectives
Design challenges
- 01
Build a dedicated dataset for five visually similar classes.
- 02
Train and version models with a reproducible workflow.
- 03
Make inference usable on images, video, or webcam.
Visual evidence
The project in view
Architecture
Technical flow
Collection and labeling → YOLO dataset → YOLOv5 training → validation → versioned weights → image or video inference
Scroll through the flow
Approach
Working steps
- 01
Assemble and label the dataset with dedicated tools.
- 02
Train two model versions and retain their weights.
- 03
Compare labels and predictions on the published validation batches.
Public repository
What the sources document
- Two model versions, their weights, and their confusion matrices are published in the repository.
- Validation examples allow annotations and predictions to be compared directly.