June 16, 2020
Computer Sciences: Results of our collaborative work for Multimodal Future Localization presented at CVPR 2020
At the CVPR 2020 conference, our collaborators from the Computer Vision Group at the University of Freiburg, Germany and ourselves, we presented the results of our work on multiple future prediction.
The algorithm is able to predict that each of the pedestrians crossing the road in front of us can be in the future in three different localization (and compensating the motion of ego vehicle), each one with its own probability of occurrence (as illustrated with different heatmaps).
This month at the CVPR'20 virtual conference, our collaborators from the Computer Vision Group at the University of Freiburg, Germany and us presented the results of our work on multiple future prediction funded by IMRA Europe S.A.S.
This work is entitled “Multimodal Future Localization and Emergence Prediction for Objects in Egocentric View with a Reachability Prior” and presents a novel approach of a unified framework for predicting multiple future positions of a moving object and multiple future emergence of not yet seen objects as multimodal probability distributions, constrained by a reachability prior.
Congratulations to Osama Makansi, the main author of this work, and to all his colleagues, for this great academic achievement!
Article and related material: here (CVPR 2020 at CVF)
Video presentation: here (youtube)
Source code: here (Github)
This month at the CVPR'20 virtual conference, our collaborators from the Computer Vision Group at the University of Freiburg, Germany and us presented the results of our work on multiple future prediction funded by IMRA Europe S.A.S.
This work is entitled “Multimodal Future Localization and Emergence Prediction for Objects in Egocentric View with a Reachability Prior” and presents a novel approach of a unified framework for predicting multiple future positions of a moving object and multiple future emergence of not yet seen objects as multimodal probability distributions, constrained by a reachability prior.
Congratulations to Osama Makansi, the main author of this work, and to all his colleagues, for this great academic achievement!
Article and related material: here (CVPR 2020 at CVF)
Video presentation: here (youtube)
Source code: here (Github)