Animesh-Kr/Human-Eye-Disease-Prediction: Initial Research Release - Hybrid CNN-Transformer Platform

Read full text Summarise with VIRAT
Authors: Animesh A Kumar
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

πŸ‘οΈ Release v1.0.0 β€” Official Research Implementation This is the initial official release of the Human Eye Disease Prediction platform, developed as part of the MSc Advanced Computer Science program at Newcastle University (2025–26). πŸ”¬ Research Highlights Hybrid Architecture: EfficientNetV2L + 4Γ— Multi-Head Attention Transformer + XGBoost. Statistical Rigor: Full 5-seed validation (Mean Accuracy: 95.43% Β± 0.27%). Safety Layer: Integrated OOD detection (Mahalanobis) and Uncertainty Quantization (MC Dropout). Explainability: Support for Grad-CAM, SHAP, and Attention Map visualizations. πŸ“š Research Publication & Citation Preprint: [arXiv:2607.09809 [eess.IV]](https://arxiv.org/abs/2607.09809) Citation Link: [https://doi.org/10.48550/arXiv.2607.09809](https://doi.org/10.48550/arXiv.2607.09809) πŸ“¦ Model Assets & Weights Due to the large file size (~2.07 GB), the high-fidelity weights and calibration tensors are hosted externally: Model Weights: [huggingface.co/animeshakr/oct-retinal-weights](https://huggingface.co/animeshakr/oct-retinal-weights) Live Demo: [huggingface.co/spaces/animeshakr/oct-retinal-ai](https://huggingface.co/spaces/animeshakr/oct-retinal-ai) πŸ“œ Attribution Author: Animesh Kumar ORCID: [0009-0003-0608-7004](https://orcid.org/0009-0003-0608-7004) Institution: Newcastle University, UK Full Changelog: https://github.com/Animesh-Kr/Human-Eye-Disease-Prediction/commits/v1.0.0

Fields of Study & Subjects

Retinal Imaging and Analysis Advanced Neural Network Applications Ophthalmology and Visual Health Research

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN