Smart Skin Cancer Detection



  • Smart Skin Cancer Detection classifies pictures taken from camera or locally stored images into the following categories: Melanoma, Nevus and Seborrheic Keratosis.

  • Based on a local running Convolutional Neural Network a prediction is listed on the screen, without the need of an internet connect.

  • All operations are handled locally on your mobile device. Your skin pictures are handled private. No network connection needed.

  • Also a history of your skin is kept giving a medical professional the option to review the development by forwarding a close up from the skin history view viaemail.

  • Additionally a periodical reminder can be set by the app in your calendar.

  • The Help text will give you a deep explantation on how to work with the app.


  • Click on Icon to zoom in.


    SmartSkin launch screen Scan your skin Take picture Focus on interesting parts Result from the Convolutional Neural Network Save the scan to the history Review your scans in the skin history Help


    Disclaimer

  • The app does neither replace a dermatologist visit nor can the result seen as a proper diagnosis however the app will raise your awareness of your skin and lead you to better health !

  • The app saves data localy on your mobile only.

  • Smart Skin keeps history of your skin. Sharing the images makes very much sense but its up to you to use this functionality.



  • Architecture



  • Smart Skin Cancer Detection is based on the TensorFlow light framework.

  • For the app the MobileNet 1.0.224 CNN was retrained with the free available ISIC datasets.

  • If you are interested on how the CNN was manipulated and trained please don't hesitate to drop me an Email