ECCV 2022 - Wednesday

Kaan tells us they are already working on an extension to the project focusing on the interaction between objects. “ By modeling the interaction between objects, we think it’ll improve the performance of each trajectory or each future prediction for the objects, ” he reveals. “ We’ll use intuition from trajectory prediction literature, like GNNs or attention type of connections, to model this. ” This work is Kaan’s second poster, but the first in person, having presented virtually at ICCV last year. How does it feel to have a poster accepted? “ Yeah, it’s really cool! ” he smiles. “ Especially having your poster accepted at a top-tier conference like ECCV – it’s great! ” Perhaps an oral at CVPR next? “ We hope so! ” he laughs. Kaan works mainly on future prediction, but the representation changes depending on the project. He also works on trajectory prediction and has worked on both video prediction and bird’s -eye view prediction as well. Previously, Fatma was a PhD student in Germany working with Andreas Geiger , then a postdoc at the University of Oxford , working with Andrea Vedaldi and Andrew Zisserman . As an Assistant Professor in Istanbul, her team focuses mainly on self-driving, and most of their recent 24 Poster Presentation