DualCross: Cross-Modality Cross-Domain Adaptation for Monocular BEV Perception

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Man, Yunze, Gui, Liang-Yan, Wang, Yu-Xiong
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917690954743808
author Man, Yunze
Gui, Liang-Yan
Wang, Yu-Xiong
author_facet Man, Yunze
Gui, Liang-Yan
Wang, Yu-Xiong
contents Closing the domain gap between training and deployment and incorporating multiple sensor modalities are two challenging yet critical topics for self-driving. Existing work only focuses on single one of the above topics, overlooking the simultaneous domain and modality shift which pervasively exists in real-world scenarios. A model trained with multi-sensor data collected in Europe may need to run in Asia with a subset of input sensors available. In this work, we propose DualCross, a cross-modality cross-domain adaptation framework to facilitate the learning of a more robust monocular bird's-eye-view (BEV) perception model, which transfers the point cloud knowledge from a LiDAR sensor in one domain during the training phase to the camera-only testing scenario in a different domain. This work results in the first open analysis of cross-domain cross-sensor perception and adaptation for monocular 3D tasks in the wild. We benchmark our approach on large-scale datasets under a wide range of domain shifts and show state-of-the-art results against various baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2305_03724
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle DualCross: Cross-Modality Cross-Domain Adaptation for Monocular BEV Perception
Man, Yunze
Gui, Liang-Yan
Wang, Yu-Xiong
Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
Closing the domain gap between training and deployment and incorporating multiple sensor modalities are two challenging yet critical topics for self-driving. Existing work only focuses on single one of the above topics, overlooking the simultaneous domain and modality shift which pervasively exists in real-world scenarios. A model trained with multi-sensor data collected in Europe may need to run in Asia with a subset of input sensors available. In this work, we propose DualCross, a cross-modality cross-domain adaptation framework to facilitate the learning of a more robust monocular bird's-eye-view (BEV) perception model, which transfers the point cloud knowledge from a LiDAR sensor in one domain during the training phase to the camera-only testing scenario in a different domain. This work results in the first open analysis of cross-domain cross-sensor perception and adaptation for monocular 3D tasks in the wild. We benchmark our approach on large-scale datasets under a wide range of domain shifts and show state-of-the-art results against various baselines.
title DualCross: Cross-Modality Cross-Domain Adaptation for Monocular BEV Perception
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2305.03724