價格：免費
更新日期：2017-01-23
檔案大小：12M
目前版本：5.0.1
版本需求：Android 4.1 以上版本
官方網站：mailto:alexandrepdumont@gmail.com
Email：https://mlechecker1203201620161203103454.azurewebsites.net/
MoleChecker is an app that uses neural network machine learning to produce a discrimination between different skin conditions. It is not meant to serve as a diagnostic application, merely as a novelty application. The UI is simple bt I think effective. The app is completely user generated. A user may submit either testing or training images. If the user submits a training image, he or she must also submit an associated condition. This will train the neural network with back-propagation running on the server side. A user may also decide to submit a testing image, in which case the neural network will try to work out what are the most likely conditions for that image.
It is very important when submitting training data that the flash of the phone be turned ON. This app works off of diffuse reflectance. The amount of light reflected off the skin and into the camera is a percentage of the out-going light from the flash. Based on how much light is returned to the camera can give us an idea of chromophore concentration for that particular spot, and varying chromophore concentrations gives us some insight on the physiological properties of that spot.
