Detect closer surfaces that can be seen: new modeling and evaluation in cross-domain 3D object detection
Detect closer surfaces that can be seen: new modeling and evaluation in cross-domain 3D object detection
The performance of domain adaptation technologies has not yet reached an ideal level in the current 3D object detection field for autonomous driving, which is mainly due to significant differences in the size of vehicles, as well as the environments they operate in when applied across domains. These factors together hinder the effective transfer and application of knowledge learned from specific datasets. Since the existing evaluation metrics are initially designed for evaluation on a single domain by calculating the 2D or 3D overlap between the prediction and ground-truth bounding boxes, they often suffer from the overfitting problem caused by the size differences among datasets. This raises a fundamental question related to the evaluation of the 3D object detection models’ cross-domain performance: Do we really need models to maintain excellent performance in their original 3D bounding boxes after being applied across domains? From a practical application perspective, one of our main focuses is actually on preventing collisions between vehicles and other obstacles, especially in cross-domain scenarios where correctly predicting the size of vehicles is much more difficult. In other words, as long as a model can accurately identify the closest surfaces to the ego vehicle, it is sufficient to effectively avoid obstacles. In this paper, we propose two metrics to measure 3D object detection models’ ability of detecting the closer surfaces to the sensor on the ego vehicle, which can be used to evaluate their cross-domain performance more comprehensively and reasonably. Furthermore, we propose a refinement head, named EdgeHead, to guide models to focus more on the learnable closer surfaces, which can greatly improve the cross-domain performance of existing models not only under our new metrics, but even also under the original BEV/3D metrics. Our code is available at https://github.com/Galaxy-ZRX/EdgeHead.
65-72
Zhang, Ruixiao
fc3c4eb9-b692-4ab3-8056-030cb6731fc5
Wu, Yihong
2876bede-25f1-47a5-9e08-b98be99b2d31
Lee, Juheon
cd382ebf-0bcc-47b8-a60d-68c6540d31bb
Cai, Xiaohao
de483445-45e9-4b21-a4e8-b0427fc72cee
Prugel-Bennett, Adam
b107a151-1751-4d8b-b8db-2c395ac4e14e
16 October 2024
Zhang, Ruixiao
fc3c4eb9-b692-4ab3-8056-030cb6731fc5
Wu, Yihong
2876bede-25f1-47a5-9e08-b98be99b2d31
Lee, Juheon
cd382ebf-0bcc-47b8-a60d-68c6540d31bb
Cai, Xiaohao
de483445-45e9-4b21-a4e8-b0427fc72cee
Prugel-Bennett, Adam
b107a151-1751-4d8b-b8db-2c395ac4e14e
Zhang, Ruixiao, Wu, Yihong, Lee, Juheon, Cai, Xiaohao and Prugel-Bennett, Adam
(2024)
Detect closer surfaces that can be seen: new modeling and evaluation in cross-domain 3D object detection.
Endriss, Ulle, Melo, Francisco S., Bach, Kerstin, Bugarin-Diz, Alberto, Alonso-Moral, Jose M., Barro, Senen and Heintz, Fredrik
(eds.)
In ECAI 2024 - 27th European Conference on Artificial Intelligence, Including 13th Conference on Prestigious Applications of Intelligent Systems, PAIS 2024, Proceedings.
vol. 392,
IOS Press.
.
(doi:10.3233/FAIA240472).
Record type:
Conference or Workshop Item
(Paper)
Abstract
The performance of domain adaptation technologies has not yet reached an ideal level in the current 3D object detection field for autonomous driving, which is mainly due to significant differences in the size of vehicles, as well as the environments they operate in when applied across domains. These factors together hinder the effective transfer and application of knowledge learned from specific datasets. Since the existing evaluation metrics are initially designed for evaluation on a single domain by calculating the 2D or 3D overlap between the prediction and ground-truth bounding boxes, they often suffer from the overfitting problem caused by the size differences among datasets. This raises a fundamental question related to the evaluation of the 3D object detection models’ cross-domain performance: Do we really need models to maintain excellent performance in their original 3D bounding boxes after being applied across domains? From a practical application perspective, one of our main focuses is actually on preventing collisions between vehicles and other obstacles, especially in cross-domain scenarios where correctly predicting the size of vehicles is much more difficult. In other words, as long as a model can accurately identify the closest surfaces to the ego vehicle, it is sufficient to effectively avoid obstacles. In this paper, we propose two metrics to measure 3D object detection models’ ability of detecting the closer surfaces to the sensor on the ego vehicle, which can be used to evaluate their cross-domain performance more comprehensively and reasonably. Furthermore, we propose a refinement head, named EdgeHead, to guide models to focus more on the learnable closer surfaces, which can greatly improve the cross-domain performance of existing models not only under our new metrics, but even also under the original BEV/3D metrics. Our code is available at https://github.com/Galaxy-ZRX/EdgeHead.
Text
FAIA-392-FAIA240472
- Version of Record
More information
Published date: 16 October 2024
Venue - Dates:
27th European Conference on Artificial Intelligence, ECAI 2024, , Santiago de Compostela, Spain, 2024-10-19 - 2024-10-24
Identifiers
Local EPrints ID: 502154
URI: http://eprints.soton.ac.uk/id/eprint/502154
ISSN: 0922-6389
PURE UUID: 6a28e33f-0bef-4e8d-ba71-e52d5ac431b7
Catalogue record
Date deposited: 17 Jun 2025 16:50
Last modified: 22 Aug 2025 02:29
Export record
Altmetrics
Contributors
Author:
Ruixiao Zhang
Author:
Yihong Wu
Author:
Juheon Lee
Author:
Xiaohao Cai
Author:
Adam Prugel-Bennett
Editor:
Ulle Endriss
Editor:
Francisco S. Melo
Editor:
Kerstin Bach
Editor:
Alberto Bugarin-Diz
Editor:
Jose M. Alonso-Moral
Editor:
Senen Barro
Editor:
Fredrik Heintz
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics