C-BRIC Purdue Students Present at CVPR 2022

C-BRIC researchers Isha Garg and Amitangshu Mukherjee of Purdue University joined IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) to present their work on "Encoding Hierarchical Information in Neural Networks helps in Subpopulation Shift." Their work focuses on the problem of subpopulation shift, a core problem in computer vision. This problem arises when a network trained on one set of sub-populations of labels (such as breeds of dogs) does not generalize to unseen sub-populations under the same label. To tackle this problem, the authors developed a novel training algorithm for neural networks to incorporate the knowledge of hierarchies to make them more robust to the problem of subpopulation shift. 

 

CVPR is "the premier annual computer vision event" where participants were invited to attend different workshops, tutorials, and a field exhibition of demonstrations of the latest technologies. Garg and Mukherjee presented their work in the Ninth Workshop on Fine-Grained Visual Categorization at the event. The CVPR was held in person June 19-24, 2022, at the New Orleans Ernest N. Morial Convention Center in New Orleans, Louisiana, and also held virtually. CVPR allowed attendees to network with leaders in academia and industry and explore new technologies from companies such as Meta, Tesla, and Apple at the expo held during the conference.