Efficient Multi-View 3D Object Detection by Dynamic Token Selection and Fine-Tuning

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) $\mathtt{\text{ToC3D}}$ 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 $\mathtt{\text{ToC3D}}$, 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 $48\mathrm{\%}$ … $55\mathrm{\%}$, inference latency (on an $\mathtt{\text{NVIDIA-GV100}}$ GPU) by $9\mathrm{\%}$ … $25\mathrm{\%}$, while still improving mean average precision by $1.0\mathrm{\%}$ … $2.8\mathrm{\%}$ absolute and NuScenes detection score by $0.4\mathrm{\%}$ … $1.2\mathrm{\%}$ absolute compared to so-far SOTA $\mathtt{\text{ToC3D}}$.