| Full name | 3DJCG (Captioning) (VoteNet + Feature-Enhancement + Transformer-Based-Head) |
| Description | Joint Training
We use the VoteNet backbone for detection. |
| Publication title | 3DJCG: A Unified Framework for Joint Dense Captioning and Visual Grounding on 3D Point Clouds |
| Publication authors | Daigang Cai, Lichen Zhao, Jing Zhang†, Lu Sheng, Dong Xu |
| Publication venue | CVPR2022 Oral |
| Publication URL | https://openaccess.thecvf.com/content/CVPR2022/papers/Cai_3DJCG_A_Unified_Framework_for_Joint_Dense_Captioning_and_Visual_CVPR_2022_paper.pdf |
| Input Data Types | Uses XYZ coordinates,Uses Multiview Image Features,Uses Normal Vectors |
| Programming language(s) | Python With Cuda |
| Hardware | GeForce RTX 2080 Ti, 11GB RAM |
| Source code or download URL | https://github.com/zlccccc/3DJCG |
| Submission creation date | 12 Sep, 2022 |
| Last edited | 13 Sep, 2022 |