Computer Vision News - May 2019

of annotated CT scans. It is also important to include all kind of pathologies. We have developed a proprietary medical data augmentation system: modified CT scans look like real ones, even though the shown anatomy is very different. Validated annotation requires a long work, even when we use semi- automated tools. The solution must be validated with the highest accuracy and objectivity. There must be a test set which does not participate in the training and which is never used by the R&D engineer to build the algorithm. This test set must be varied and large enough to be relevant and credible, at least as large as the training set. Read about more projects in medical segmentation Project 15 Computer Vision News A project by RSIP Vision Take us along for your next Deep Learning project! Request a call here