Showing 1–1 of 1 results for author: Voodarla, M
-
S-BEV: Semantic Birds-Eye View Representation for Weather and Lighting Invariant 3-DoF Localization
Authors:
Mokshith Voodarla,
Shubham Shrivastava,
Sagar Manglani,
Ankit Vora,
Siddharth Agarwal,
Punarjay Chakravarty
Abstract:
We describe a light-weight, weather and lighting invariant, Semantic Bird's Eye View (S-BEV) signature for vision-based vehicle re-localization. A topological map of S-BEV signatures is created during the first traversal of the route, which are used for coarse localization in subsequent route traversal. A fine-grained localizer is then trained to output the global 3-DoF pose of the vehicle using i…
▽ More
We describe a light-weight, weather and lighting invariant, Semantic Bird's Eye View (S-BEV) signature for vision-based vehicle re-localization. A topological map of S-BEV signatures is created during the first traversal of the route, which are used for coarse localization in subsequent route traversal. A fine-grained localizer is then trained to output the global 3-DoF pose of the vehicle using its S-BEV and its coarse localization. We conduct experiments on vKITTI2 virtual dataset and show the potential of the S-BEV to be robust to weather and lighting. We also demonstrate results with 2 vehicles on a 22 km long highway route in the Ford AV dataset.
△ Less
Submitted 23 January, 2021;
originally announced January 2021.
