X
Hugging Face
BlogSkin-AI
Published:
Skin AI is an educational dermatology assistant that classifies a submitted skin image into one of 22 condition categories. I built it to explore how image classification could make preliminary health information more accessible when specialist care is difficult to reach.
The application uses a ConvNeXt model and returns class confidence scores with a simple risk summary. A Gradio interface makes the model accessible in a browser, while ONNX Runtime provides an optimized inference option. The project also includes Docker configuration for reproducible deployment.
Features
- classification across 22 skin condition categories;
- ConvNeXt image classification model;
- confidence scores and risk categories;
- PyTorch and ONNX inference;
- Gradio web interface;
- Docker deployment support.
Supported categories
The categories include acne, actinic keratosis, benign tumors, bullous diseases, candidiasis, drug eruptions, eczema, infestations and bites, lichen, lupus, moles, psoriasis, rosacea, seborrheic keratoses, skin cancer, sun damage, tinea, vascular tumors, vasculitis, vitiligo, warts, and unknown or normal cases.
Important limitation
Skin AI is an educational and informational project. Its output is not a medical diagnosis and should not replace evaluation by a qualified dermatologist.