search

3d-vista / 3D-VisTA

Official implementation of ICCV 2023 paper "3D-VisTA: Pre-trained Transformer for 3D Vision and Text Alignment"
https://3d-vista.github.io
MIT License
189 stars 11 forks source link

ScanScribe dataset #10

Closed jiaqchen closed 1 year ago

jiaqchen commented 1 year ago

Hello! Cool work! Will there be more data released? I think there's currently only 219 unique scenes from 3DSSG scenes right?

zhuziyu-edward commented 1 year ago

For ScanNet scenes, texts are from ScanQA, ScanRefer and ReferIt3D training set.