Computer Vision News - November 2022

48 ECCV Workshop chestX-rayimagesandhistopathology images. The third prize was awarded to Romain Mormont et al. , from the University of Liège, France, for the work entitled “Relieving pixel- wise labelling effort for pathology image segmentation with self-training”. In order to combine both exhaustively annotated and very sparsely-annotated images, the authors proposed a self-training based approach for image segmentation tasks, The work presented by Dwarikanath Mahapatra, from the Inception Institute of Artificial Intelligence, UAE, came in second place, proposing an approach to improve the generalization capability of classification models for medical images. With graph neural networks, unsupervised learning, disentangled semantic and domain invariant structural features, the authors showed better performance across distribution shifts, in two types of data: BEST ECC