Computer Vision News - July 2022

BEST OF CVPR 38 CVPR Workshop outstanding performances by utilizing unlabeled data. Pablo Arbelaez from Universidad de los Andes , gave a wonderful presentation on how to build fully autonomous surgical system to facilitate surgeons . Especially, he talked about the importance of combining both long-term (surgical phases and steps) and short-term (instrument type, localization and atomic actions in frames) understanding for improving holistic surgical scene understanding. The PSI- AVA dataset offers a great opportunity for future researchers to delve deeper into this direction. Mert Sabuncu from Cornell Tech is the last featured speaker for this session. His talk on “Neural Encoding Models” revealed great insights into the correlation between attention and eye gaze, which explains why visual attention plays a vital role in neural encoding. Further, Mert introduced his privacy regulations, since AI models trained from small data sets are usually not accurate and generalizable. This new training paradigm enables multiple medical institutions to train a model collaboratively without data sharing. In this unique learning regime, Xiaoxiao’s team has investigated novel optimization and learning schemes to tackle data heterogeneity, reduce dependency on data labeling, and adapt FL to different applications.  Ismail Ben Ayed from École de Technologie Supérieur (ETS) , introduced how to fully leverage unlabeled data to enhancemodel generalization inabreadthof real scenarios and applications . Especially, he talked about few-shot learning, unsupervised domain adaptation ad test-time adaptation as a few representative methods. He further introduced a series of latest works that use structure-driven / knowledge- driven / invariance / multi-modal priors to improve the above methods which achieve