Champlain Hackathon 2025 - Trash Classifier 🗑️🤖

Welcome to our Champlain Hackathon 2025 project! This app uses deep learning to classify images of waste into organic or recycling categories. Perfect for building smarter recycling solutions! 🌱♻️

---

Features ✨

• Image Classification: Upload or provide a URL to an image of trash.
• Real-time Predictions: Get results instantly via a Flask API.
• Confidence Scores: See how confident the model is about its prediction.
• Easy Integration: Use the classify_image() function directly in Python.

---

How It Works 🛠️

1. The model is a custom LeNet CNN built with PyTorch.
2. Images are resized to 128x128, normalized, and passed through the model.
3. The model outputs probabilities for each class: organic or recycling.
4. Results include predicted class and confidence.

---

Getting Started 🚀

Requirements

• Python 3.8+
• PyTorch
• torchvision
• Flask
• Pillow
• requests

Install Dependencies

pip install torch torchvision flask pillow requests flask-cors

Running the App

python backend.py

Open your browser at http://localhost:65535 to access the web interface.

API Usage

Send a POST request to /predict with:

• image (file upload)
• url (image URL)
• image_data (base64 string)

Example using Python requests:

import requests

url = 'http://localhost:65535/predict'
files = {'image': open('trash.jpg', 'rb')}
response = requests.post(url, files=files)
print(response.json())

---

Direct Python Usage 🐍

from backend import classify_image

result = classify_image('trash.jpg')
print(result)

---

Classes

• organic 🌿
• recycling ♻️

---

Have fun sorting trash with AI! 😎