InfoCycle ♻️

An AI Recycle Classifier

An AI-powered tool that identifies and classifies waste materials such as plastic, glass, paper, and metal using image recognition.
The goal of InfoCycle is to help users determine whether an item is recyclable with just a single image.

In the future, InfoCycle will (hopefully) be available as a simple web app where users can upload a photo and instantly receive a classification result.
If not deployed as a web app, it will include a standalone Python script (predict.py) that takes an image file and outputs the detected material type and recyclability!

Features (Planned)

• Classify uploaded images into material categories (plastic, paper, glass, metal, etc.)
• Determine recyclability directly from an image
• Fast and lightweight inference with a pre-trained model
• Option to run locally using predict.py
• Future web integration with an upload based UI
• Powered by a ResNet-18 deep learning model trained in PyTorch

Current Progress

• Project structure created and organized
• Environment setup with GPU enabled PyTorch (for me only)
• (WIP) Developing and training the ResNet-18 model (train_model.ipynb)

Future Plans

• Complete model training and release downloadable model
• Add predict.py for quick local image testing
• Display recyclability confidence score and category breakdown
• Develop and deploy web app

Project Structure

InfoCycle/  
    model/                Trained model weights (to be added)  
    notebooks/            Model training and experimentation  
        ── train_model.ipynb  
    src/                  Future: predict.py and web app scripts