Project for a neuro-symbolic Image classifier utilizing a neural network and a ruleset induced from a decision tree to classfy images from the CIFAR-10 dataset in an explainable and efficient manner.
📂 project_root │
├── 📂 models # Pre-trained models for image classification
│ ├── 📄 feature_recognition_cnn.onnx # CNN model in ONNX format
│ ├── 📄 feature_recognition_cnn.pth # CNN model in PyTorch format
│ ├── 📄 neuro_symbolic_classifier.pkl # Pickled hybrid classifier
│
├── 📂 notebooks # Jupyter notebooks for training and inference
│ ├── 📄 Neurosymbolic_App.ipynb # Notebook for running the live demo app
│ ├── 📄 Neurosymbolic_Image_Classifier.ipynb # Notebook for training/testing the classifier
│
├── 📄 .gitignore # gitignore file for handling external files and directories
├── 📄 .neuro_symbolic_classifier_streamlit.py # Python file for running the Streamlit application
├── 📄 requirements.txt # Environment details necessary to run the experiments
├── 📄 README.md # Project documentation and instructions
To run the Streamlit Demo simply click the link here.
Or if you would rather see the Gradio Demo on Google Colab, then click the link below.
To run the Gradio App Demo simply run this notebook.