fruit-detection-model
Custom-trained YOLOv8 model that detects apples, bananas, oranges, grapes and strawberries in real time.
- Language
- Python
What it does
Trained on 7,000+ images, semi-automatically labeled with YOLO-World. Reports 91% mAP50 on the validation set.
Two modes: a large model for GPU servers, and a small model at 320 px input that runs at 12–20 FPS on a laptop CPU.
Why I built it
I wanted to take an object detector through the whole path: building a labeled dataset, training, and trading accuracy for speed so it runs on ordinary hardware.
Install and use
git clone https://github.com/apurboislam/fruit-detection-model.git
cd fruit-detection-model
pip install ultralytics opencv-python
# download best.pt / best_small.pt from the v1.0 releasepython main.py # large model, needs a GPU
python main_small.py # small model, runs on CPULive demo: not shipped yet
A browser demo of fruit-detection-model will live here. For now, see the repository.