CVPR Daily - Wednesday

questions. This excellent work won the Best Paper Award at the workshop. Timo Kaiser from Leibniz Universitat Hannover discussed their method to identify and compensate label ambiguities as and label in semantic segmentation. Zhihao Zhao from University of California, Berkeley, presented a novel RGB-D dataset, called Digital Twin Tracking Dataset to help researchers develop future object tracking methods and analyze new challenges. Shuyu Miao from Ant Group introduced their winning solution of the DataCV challenge where they proposed a dynamic regression model to capture the relationship between the shifts in distribution and model accuracy. Jihun Yoon from AI Dev. Group, Hutom, introduced a few attempts in reducing frame redundancies to enable efficient learning in instance segmentation. Aboli Marathe from Carnegie Mellon University presented WEDGE (WEather images by DALL-E GEneration), a synthetic dataset generated with a vision-language generative model via prompting. It simulates 16 extreme weather conditions and supports research in the tasks of weather classification and 2D object detection. I met Andrew Ng afterwards and discussed with him the perspectives of the workshop. We can invite Andrew to give a talk in the 2024 edition, perhaps with a different name, e.g., Data-centric CV Workshop. Through this initiative, we hope to further grow and foster this community on computer vision data analysis and hope to join force with the model community and the broader machine learning family. 19 DAILY CVPR Wednesday QCVML - Quantum CV …