How it works
Three steps from photo to insight.
1 · Upload
Drag & drop or select a clear, well-lit photo of the lesion. JPG or PNG, up to 10MB.
2 · Analyze
The image is resized to 224×224, normalized, and passed through a CNN trained on ISIC dermoscopic data.
3 · Review
Get a Benign/Malignant label, confidence score, Grad-CAM overlay, and downloadable PDF report.
Under the hood
- • Preprocessing: resize 224×224, channel-wise mean/std normalization.
- • Backbone: ResNet-50 / MobileNetV2 fine-tuned on ISIC 2019.
- • Output: softmax over 7 lesion classes, mapped to benign/malignant.
- • Explainability: Grad-CAM activation heatmap overlaid on the original.