Computer Vision News - January 2022

31 Most Read in 2021 Swin Transformer: Hierarchical Vision Transformer Using Shifted Windows - Best Paper ICCV21 Out of 1,500 papers to be presented at ICCV 2021 , we choose the one which later won the Best Paper Award. Believe it or not, our readers know that we are often very lucky: this has happened many times. Co-author Han Hu of Microsoft Research Asia told us about this work proposing a new architecture based on Swin Transformer that performs much better than CNN backbones. Read It Here Class Imbalance in Classification Tasks - with code! Datasets are called imbalanced when the distribution of examples across classes varies heavily, which oftenhappens incomputer visionand medical imaging applications. It can be due to incorrect data sampling or an inherent property of the domain. In most cases, it is advised to address this issue using specific tools and ad hoc libraries. Our Marica Muffoletto gives you precious tips, with code! Read It Here Tool Analysis - InnerEye by Microsoft: Democratizing Medical Imaging AI - with code! Microsoft Research Cambridge recently developed a tool, aiming at “democratize medical imaging AI”. Sounds like a good bet, and apparently it is, if readers adored this report - with code! - written by our Ioannis Valasakis . Using state of the art and automatic deep learning on medical imaging on the cloud, InnerEye provides a safe, coherent workflow with great data management. Read It Here