sangavi-a/FRUIT-DETECTION

metadata title: Fruit Object Detection emoji: 🍎 colorFrom: green colorTo: blue sdk: docker app_file: app.py pinned: false 🍎🍌🍊 Fruit Object Detection using YOLOv8 This project implements fruit detection (Apple, Banana, Orange) using Ultralytics YOLOv8. The dataset contains images of three classes and was trained, validated, and tested using custom augmentation. The final model is deployed with a Streamlit App and hosted on AWS/GCP for real-time inference.

📌 Project Overview This project identifies three fruit classes:

Apple (class 0) Banana (class 1) Orange (class 2) 📁 Project Structure project/ ├── yolodata/ │ ├── images/ │ │ ├── train/ │ │ ├── valid/ │ │ ├── test/ │ ├── labels/ │ │ ├── train/ │ │ ├── valid/ │ │ ├── test/ ├── fruit.pt ├── fruitstreamlit.py └── data.yaml

🚀 Model Summary Model: YOLOv8s Epochs: 80 Device: Google Colab GPU Early stopping enabled Augmentation applied: Horizontal Flip Brightness/Contrast Rotation Color Jitter Motion/Motion Blur Random Shadows Number of classes = 3 apple banana orange 📊 Final Evaluation Metrics Your final training metrics (real values):

Per-Class Scores Class Precision Recall F1-Score Apple 0.91114 0.99999 0.95351 Banana 0.97397 0.93333 0.95322 Orange 0.95614 0.98276 0.96927 Mean Scores Metric Value mAP50 0.97789 mAP50-95 0.83741 🔲 Confusion Matrix Your final confusion matrix: [[54 1 0 10] [ 0 57 0 6] [ 0 0 57 3] [ 0 2 1 0]]

Author SANGAVI