Re_3DVG proposes a method for fragmented point cloud scenarios. Utilizing instance segmentation and transformer models, our approach offers a potent mechanism for establishing robust correspondences between text queries and object instances within the shared visible range.
The fragmented 3DVG dataset can be downloaded from:
Google Drive
The test dataset can be downloaded from:
Google Drive
The weight module files can be downloaded from:
Google Drive
We also provide weight files in Chinese from：
Google Drive
Our project has been tested on torch=1.13 cuda=11.7 python=3.7.5
You can use the following scripts for the train: T_V_en_base_with_F_4L_e4_BERT.py
You can use the example demo_eval_iou.py to perform RGB images, depth images, and text descriptions.
See the experiment directory.
We appreciate the open-source of the following projects: MVT-3DVG , OFA, UNINEXT, GroundingDINO, ScanRefer, EDA, Mask_RCNN, and ScanNet.