The ORO Health Data Science skin-in-the-game test
Hello candidate! We are super excited that you have been selected and interested to take on our technical challenge! This document is intended for an image classification problem with 3 skin disease types. Description: In this experiment, you are expected to predict the type of skin disease seen on an image. The classes are “acne”, “herpes simplex”, and “lichen planus”. A classification report will be used to measure the model’s success. Keep in mind that the given dataset is a subset of all data and your model will be tested against a separate set of images with the same disease types. This is meant for you to demonstrate your talent. You do NOT need to create the perfect model. It is for you to decide what is enough in terms of code, documentation, etc. to demonstrate your talent.
In the submitted file you should add, 1- Documentation of experiment including hardware, software, CPU/GPU runtime information as well as a summary of the approach, limitations, and possible improvements that you would do if you had more time and computational power. 2- Code snippet that we can run and replicate your results, and a final model file that we can use to predict the classes of test images. 3- Classification Report
112 images in 3 classes
Acne: 40 images Herpes Simplex: 16 images Lichen Planus: 56 images
The images are taken from http://www.danderm-pdv.is.kkh.dk/atlas/index.html with their permission to use for this technical test.