Anyone who ever tried to do image recognition knows that the key to success is not actual neural net but well-labeled dataset. Creating it is very time consuming and expensive.
Enter Friendly’s Labeling Engine. It automatically generates all XML annotations for you, letting you focus on what you love – data science. We can generate automatic XML annotations for all darknet users (Yolo app): https://pjreddie.com/darknet/yolo/. Contrarily to what you think, this open source technology is used for a lot of legitimate goals and by a lot of legitimate companies (Philips is an example).
We can also automatically label a set of radiology images. You label the first 10 for us, then we use unsupervised machine learning and other techniques to automatically label the rest. This will save you 100K because otherwise it has to be labelled by the most expensive resource in the hospital – radiologist.