Computer Vision News - July 2020

3 Summary 15 MONAI The project came together in about 4-5 months and it was well received, and since the very start many leaders in the community joined. This includes Stephen Aylward , the head of the advisory board and Senior Director of Strategic Initiatives at Kitware , and Jorge Cardoso , Senior Lecturer in Artificial Medical Intelligence at King’s College London , who we also had the opportunity to talk to. The core group is made of experts from KCL, DKFZ, Stanford, CCDS, TUM, CAS, Kitware & Nvidia who want to share their domain specific knowledge in Medical Imaging towards the achievement of a common goal. The other members of the Advisory board: Sebastien Ourselin , Klaus Maier-Hein , Jayashree Kalpathy- Cramer , Daniel Rubin , Kevin Zhou , Nassir Navab and Andrew Feng . Introduction MO NAI is PyTorch-based and can be installed through pip with either of the two commands below: At themoment, this tool offers some foundational components that can be integrated intoPytorchprograms, amongwhichwe find: several datasets, savers andwriters (e.g. Nifty, PNG), inferers, losses, visualization tools, networks, metrics and transforms. These can bemanaged by specific engines (SupervisedTrainer, Supervised Evaluator), event handlers to check ongoing training, and metrics (MeanDice, ROCAUC). These are all brought together within comprehensive examples that the authors have intentionally built to show how these elements can be combined to get a pipeline for segmentation and classification tasks, GANs & AutoEncoders and Federated learning. pip install monai pip install git +https://github.com/Project-MONAI/MONAI#egg =MONAI Stephen Aylward