DVPE: Divided View Position Embedding for Multi-View 3D Object Detection
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866913443703947264 |
|---|---|
| author | Wang, Jiasen Li, Zhenglin Sun, Ke Liu, Xianyuan Zhou, Yang |
| author_facet | Wang, Jiasen Li, Zhenglin Sun, Ke Liu, Xianyuan Zhou, Yang |
| contents | Sparse query-based paradigms have achieved significant success in multi-view 3D detection for autonomous vehicles. Current research faces challenges in balancing between enlarging receptive fields and reducing interference when aggregating multi-view features. Moreover, different poses of cameras present challenges in training global attention models. To address these problems, this paper proposes a divided view method, in which features are modeled globally via the visibility crossattention mechanism, but interact only with partial features in a divided local virtual space. This effectively reduces interference from other irrelevant features and alleviates the training difficulties of the transformer by decoupling the position embedding from camera poses. Additionally, 2D historical RoI features are incorporated into the object-centric temporal modeling to utilize highlevel visual semantic information. The model is trained using a one-to-many assignment strategy to facilitate stability. Our framework, named DVPE, achieves state-of-the-art performance (57.2% mAP and 64.5% NDS) on the nuScenes test set. Codes will be available at https://github.com/dop0/DVPE. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_16955 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DVPE: Divided View Position Embedding for Multi-View 3D Object Detection Wang, Jiasen Li, Zhenglin Sun, Ke Liu, Xianyuan Zhou, Yang Computer Vision and Pattern Recognition Robotics Sparse query-based paradigms have achieved significant success in multi-view 3D detection for autonomous vehicles. Current research faces challenges in balancing between enlarging receptive fields and reducing interference when aggregating multi-view features. Moreover, different poses of cameras present challenges in training global attention models. To address these problems, this paper proposes a divided view method, in which features are modeled globally via the visibility crossattention mechanism, but interact only with partial features in a divided local virtual space. This effectively reduces interference from other irrelevant features and alleviates the training difficulties of the transformer by decoupling the position embedding from camera poses. Additionally, 2D historical RoI features are incorporated into the object-centric temporal modeling to utilize highlevel visual semantic information. The model is trained using a one-to-many assignment strategy to facilitate stability. Our framework, named DVPE, achieves state-of-the-art performance (57.2% mAP and 64.5% NDS) on the nuScenes test set. Codes will be available at https://github.com/dop0/DVPE. |
| title | DVPE: Divided View Position Embedding for Multi-View 3D Object Detection |
| topic | Computer Vision and Pattern Recognition Robotics |
| url | https://arxiv.org/abs/2407.16955 |