I am so happy that PTv3 was selected as one of the Oral papers (90/11,532) of
#CVPR2024
. PTv2 is my first paper and the first paper accepted by a top conference. It really means a lot that PTv3 became my first Oral paper! Love you, CVPR!
Our new method Point Transformer V3 (PTv3), a stride toward a simpler, faster, and stronger design of 3D transformers, is now available in ArXiv (). This is a joint effort with Li Jiang,
@PengshuaiWang
,
@zhijianliu_
,
@HengshuangZhao
Our new method Point Transformer V3 (PTv3), a stride toward a simpler, faster, and stronger design of 3D transformers, is now available in ArXiv (). This is a joint effort with Li Jiang,
@PengshuaiWang
,
@zhijianliu_
,
@HengshuangZhao
#CVPR23
Introducing Pointcept, a codebase for point cloud perception research. Pointcept is a highly potent tool, as evidenced by its ability to boost the base semseg result of SpUnet on the ScanNet val set, resulting in an impressive 75.0%+ mIoU.
GitHub:
#CVPR
Our PTv3 is accepted by CVPR24, especially thanks to efforts from our coauthors!
@PengshuaiWang
@zhijianliu_
@HengshuangZhao
. We will also have a tutorial related to Point Cloud Understanding in
@CVPR
2024, more information will be available soon!
Point Transformer V3: Simpler, Faster, Stronger
paper page:
This paper is not motivated to seek innovation within the attention mechanism. Instead, it focuses on overcoming the existing trade-offs between accuracy and efficiency within the context of
Check our new work Point Prompt Training (PPT), a novel pre-training framework for 3D multi-dataset synergistic learning. PPT attains SOTA results across over ten tasks spanning both indoor and outdoor 3D scenarios.
ArXiv:
Code:
Happy to announce our
#CVPR2023
work, Masked Scene Contrast (MSC)!
Inspired by PointContrast and MAE, we directly operate CL and MPM on the point cloud. MSC boosted SparseUNet validation result on ScanNet to 75.5%
Arxiv:
Code:
Check our new work Point Prompt Training (PPT), a novel pre-training framework for 3D multi-dataset synergistic learning. PPT attains SOTA results across over ten tasks spanning both indoor and outdoor 3D scenarios.
ArXiv:
Code:
#CVPR23
Introducing Pointcept, a codebase for point cloud perception research. Pointcept is a highly potent tool, as evidenced by its ability to boost the base semseg result of SpUnet on the ScanNet val set, resulting in an impressive 75.0%+ mIoU.
GitHub:
Garment3DGen
3D Garment Stylization and Texture Generation
We introduce Garment3DGen a new method to synthesize 3D garment assets from a base mesh given a single input image as guidance. Our proposed approach allows users to generate 3D textured clothes based on both