🧠 Virtual Lesion Analyzer
CNN vs Transformer comparison on brain tumor MRI classification.
Upload a brain MRI scan to see how a CNN (ResNet-50) and a Vision Transformer (ViT-B/16) analyze the same image differently. The tool generates four visualizations for each architecture:
- Grad-CAM / Attention Rollout — Where the model focuses
- SmoothGrad — Pixel-level sensitivity
- Focus Region — High-activation areas highlighted
⏳ Note: The first analysis may take 1–2 minutes as the models (~420MB combined) load into memory on Hugging Face's free CPU infrastructure. Subsequent analyses will be faster. This is a research demo running on donated compute — thank you for your patience.
Models: ResNet-50 (90MB) · ViT-B/16 (330MB) · Fine-tuned on 7,023 brain MRI scans
Classes: Glioma · Meningioma · No Tumor · Pituitary
Platform: arxelos.com · Built by Aryan Patel
⚠️ Research demo only. This tool is for educational and research purposes. It is not a medical diagnostic tool. Never use this for clinical decision-making. Always consult a qualified healthcare professional.