Li3DeTr: A LiDAR based 3D Detection Transformer

TL;DR

Li3DeTr is an innovative end-to-end LiDAR-based 3D detection transformer that leverages sparse convolution and deformable attention, achieving state-of-the-art results in autonomous driving datasets.

Contribution

The paper introduces Li3DeTr, a novel transformer architecture for 3D detection from LiDAR data, with a unique cross-attention block and set-to-set loss, outperforming existing methods.

Findings

Achieves 61.3% mAP and 67.6% NDS on nuScenes

Surpasses state-of-the-art methods with NMS

Knowledge distillation slightly improves performance

Abstract

Inspired by recent advances in vision transformers for object detection, we propose Li3DeTr, an end-to-end LiDAR based 3D Detection Transformer for autonomous driving, that inputs LiDAR point clouds and regresses 3D bounding boxes. The LiDAR local and global features are encoded using sparse convolution and multi-scale deformable attention respectively. In the decoder head, firstly, in the novel Li3DeTr cross-attention block, we link the LiDAR global features to 3D predictions leveraging the sparse set of object queries learnt from the data. Secondly, the object query interactions are formulated using multi-head self-attention. Finally, the decoder layer is repeated $L_{dec}$ number of times to refine the object queries. Inspired by DETR, we employ set-to-set loss to train the Li3DeTr network. Without bells and whistles, the Li3DeTr network achieves 61.3% mAP and 67.6% NDS surpassing the state-of-the-art methods with non-maximum suppression (NMS) on the nuScenes dataset and it also achieves competitive performance on the KITTI dataset. We also employ knowledge distillation (KD) using a teacher and student model that slightly improves the performance of our network.