CNN Image Sorter

This project uses a configurable Convolutional Neural Network (CNN) to sort images into categories. It uses a JSON-defined model architecture for easy configuration.

Directory Structure

.
├── categories/         # Training images, organized by category
├── model_cnn.py        # Shared CNN model class
├── model.jsonc         # Model architecture/configuration
├── requirements.txt    # Python dependencies
├── sort.py             # Script to sort images
├── suggestions/        # Output folder for sorted images
├── tosort/             # Images to be sorted
├── train.py            # Script to train the model

Installation

1. Clone the repository and navigate to the project folder.
2. Run ./install.sh

Usage

Training

Place your training images in categories/<category>/. Each subfolder is a class. Run:

python train.py

This will train the CNN and save the model to trained_model.pth.

Sorting

Place images to be sorted in the tosort/ folder. Run:

python sort.py

Images will be copied into suggestions/<predicted_category>/ based on the model's predictions.

Notes

• Images are automatically resized for training and sorting (no modification of originals).
• Model architecture is defined in model.jsonc.
• Corrupted or unreadable images are skipped automatically.

Requirements

• Python 3.8+
• See requirements.txt for required packages (PyTorch, torchvision, Pillow, etc.)