ShiroOnigami23/skin-cancer-detection

Skin Cancer Detection Models

Models Included

1. Fine-tuned Model (RECOMMENDED)

   • Validation Accuracy: 89.1%
   • Validation F1 Score: 89.23%
   • Training Accuracy: 96.4%
   • File: models/fine_tuned_model.pth
   • Size: 50.4 MB

2. Original Model

   • Validation Accuracy: 80.4%
   • Validation F1 Score: 83.95%
   • Training Accuracy: 69.4%
   • File: models/original_model.pth
   • Size: 132.3 MB

Classes

• akiec: Actinic keratoses and intraepithelial carcinoma
• bcc: Basal cell carcinoma
• bkl: Benign keratosis-like lesions
• df: Dermatofibroma
• mel: Melanoma
• nv: Melanocytic nevi
• vasc: Vascular lesions

Quick Start

python inference.py --model models/fine_tuned_model.pth --image your_image.jpg

Requirements

Install: pip install -r requirements.txt

Notes

• Use the fine-tuned model for best performance
• Images should be dermatoscopic skin lesion photos
• This is for research/educational purposes only