Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2604.13586 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910130328567808 |
|---|---|
| author | Nazir, Danish Hanna-Asaad, Antoine Görnhardt, Lucas Piewek, Jan Bagdonat, Thorsten Fingscheidt, Tim |
| author_facet | Nazir, Danish Hanna-Asaad, Antoine Görnhardt, Lucas Piewek, Jan Bagdonat, Thorsten Fingscheidt, Tim |
| contents | Existing multi-view three-dimensional (3D) object detection approaches widely adopt large-scale pre-trained vision transformer (ViT)-based foundation models as backbones, being computationally complex. To address this problem, current state-of-the-art (SOTA) \texttt{ToC3D} for efficient multi-view ViT-based 3D object detection employs ego-motion-based relevant token selection. However, there are two key limitations: (1) The fixed layer-individual token selection ratios limit computational efficiency during both training and inference. (2) Full end-to-end retraining of the ViT backbone is required for the multi-view 3D object detection method. In this work, we propose an image token compensator combined with a token selection for ViT backbones to accelerate multi-view 3D object detection. Unlike \texttt{ToC3D}, our approach enables dynamic layer-wise token selection within the ViT backbone. Furthermore, we introduce a parameter-efficient fine-tuning strategy, which trains only the proposed modules, thereby reducing the number of fine-tuned parameters from more than $300$ million (M) to only $1.6$ M. Experiments on the large-scale NuScenes dataset across three multi-view 3D object detection approaches demonstrate that our proposed method decreases computational complexity (GFLOPs) by $48\%$ ... $55\%$, inference latency (on an \texttt{NVIDIA-GV100} GPU) by $9\%$ ... $25\%$, while still improving mean average precision by $1.0\%$ ... $2.8\%$ absolute and NuScenes detection score by $0.4\%$ ... $1.2\%$ absolute compared to so-far SOTA \texttt{ToC3D}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_13586 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Efficient Multi-View 3D Object Detection by Dynamic Token Selection and Fine-Tuning Nazir, Danish Hanna-Asaad, Antoine Görnhardt, Lucas Piewek, Jan Bagdonat, Thorsten Fingscheidt, Tim Computer Vision and Pattern Recognition Existing multi-view three-dimensional (3D) object detection approaches widely adopt large-scale pre-trained vision transformer (ViT)-based foundation models as backbones, being computationally complex. To address this problem, current state-of-the-art (SOTA) \texttt{ToC3D} for efficient multi-view ViT-based 3D object detection employs ego-motion-based relevant token selection. However, there are two key limitations: (1) The fixed layer-individual token selection ratios limit computational efficiency during both training and inference. (2) Full end-to-end retraining of the ViT backbone is required for the multi-view 3D object detection method. In this work, we propose an image token compensator combined with a token selection for ViT backbones to accelerate multi-view 3D object detection. Unlike \texttt{ToC3D}, our approach enables dynamic layer-wise token selection within the ViT backbone. Furthermore, we introduce a parameter-efficient fine-tuning strategy, which trains only the proposed modules, thereby reducing the number of fine-tuned parameters from more than $300$ million (M) to only $1.6$ M. Experiments on the large-scale NuScenes dataset across three multi-view 3D object detection approaches demonstrate that our proposed method decreases computational complexity (GFLOPs) by $48\%$ ... $55\%$, inference latency (on an \texttt{NVIDIA-GV100} GPU) by $9\%$ ... $25\%$, while still improving mean average precision by $1.0\%$ ... $2.8\%$ absolute and NuScenes detection score by $0.4\%$ ... $1.2\%$ absolute compared to so-far SOTA \texttt{ToC3D}. |
| title | Efficient Multi-View 3D Object Detection by Dynamic Token Selection and Fine-Tuning |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.13586 |