The KITTI Vision Benchmark Suite

This is a single stage transformer architecture
operating directly on voxelized LiDAR points.
PointViT V1 introduces depth based virtual points,
integrates self attention with local depthwise
convolutions, and employs a multiscale voxel
feature encoder.

Number of transformer heads: 4

@misc{sood2026pointvirtualtransformer,
title={Point Virtual Transformer},
author={Veerain Sood and Bnalin and Gaurav
Pandey},
year={2026},
eprint={2602.06406},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2602.06406},
}