12 benign and malignant tumorous skin disorders
(basal cell carcinoma, squamous cell carcinoma, intraepithelial carcinoma, actinic keratosis, seborrheic keratosis, malignant melanoma, melanocytic nevus, lentigo, pyogenic granuloma, hemangioma, dermatofibroma, and wart)


Python example : test.py and roc.py
Requirement
- Python and BVLC PyCaffe (http://api.medicalphoto.org/bvlc_caffe.php)

Download
OneDrive : https://1drv.ms/u/s!AiDkEyRcgr75bVnsHhKs5PtL_5U

How to use : test.py

This script shows the diagnotics predicted by AI for 27 test images.
- Run "test.bat"

- Check report.txt - https://github.com/whria78/12dx/blob/master/Model/report.txt

How to use : roc.py

We created ROC curves and threshold-sensitivity-specificity curves using MatPlotLib (https://matplotlib.org).

- Run "roc - trained by Asan.bat" ( or "roc - trained by Asan small.bat")

- It takes a long time, 100 ~ 300 minutes depending on CPU 

- Check 12 ROC curve images and 12 threshold-sensitivity-specificity curve images.


Web Demo
We have created a website for mobile devices to demonstrate the possibility of the straightforward use of all kinds of smartphones.

The model of web demo (http://dx.medicalphoto.org) (http://modelderm.com) was trained with the Asan train dataset (15,408 images, and additional 170,000 images).