End-to-End Learning Improves Static Object Geo-Localization From Video

Mohamed Chaabane, Lionel Gueguen, Ameni Trabelsi, Ross Beveridge, Stephen O'Hara; Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2021, pp. 2063-2072

Abstract


Accurately estimating the position of static objects, such as traffic lights, from the moving camera of a self-driving car is a challenging problem. In this work, we present a system that improves the localization of static objects by jointly-optimizing the components of the system via learning. Our system is comprised of networks that perform: 1) 5DoF object pose estimation from a single image, 2) association of objects between pairs of frames, and 3) multi-object tracking to produce the final geo-localization of the static objects within the scene. We evaluate our approach using a publicly-available data set, focusing on traffic lights due to data availability. For each component, we compare against contemporary alternatives and show significantly-improved performance. We also show that the end-to-end system performance is further improved via joint-training of the constituent models.

Related Material


[pdf] [supp]
[bibtex]
@InProceedings{Chaabane_2021_WACV, author = {Chaabane, Mohamed and Gueguen, Lionel and Trabelsi, Ameni and Beveridge, Ross and O'Hara, Stephen}, title = {End-to-End Learning Improves Static Object Geo-Localization From Video}, booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)}, month = {January}, year = {2021}, pages = {2063-2072} }