Prediction Demo Release

This release includes a working prediction demo using a trained glaucoma diagnosis model.

Included in glaucoscan-v1.1.zip:

• predict.py — Script to test the trained model on new input
• example_input.csv — Sample 2048-length feature vector

📦Also archived on Zenodo:
🔗 https://doi.org/10.5281/zenodo.15567617

Initial Public Release of GlaucoScan AI

This is the first public release of the GlaucoScan.ai tool for early glaucoma detection using fundus images.

Model Overview:

• Model type: MLP trained on ORIGA dataset
• Input: Extracted image features (not raw images)
• Output: Binary classification (glaucoma vs. healthy)
• Optimized for low FLOPs (~0.02B) to support edge devices

Included Files:

• glaucoma_mlp_model.pkl – Trained model file
• scaler.joblib – StandardScaler used in preprocessing
• feature_selector.joblib – Feature selection object used during training
• training_config.pkl – Configuration file for reproducibility
• class_mapping.pkl – Maps model output to labels
• glaucoscan-v1.0.zip – Light version of the full rep with model files and code

Download counts are tracked per asset. Feel free to cite or fork the repo for your research or development use.