Summary of Sceneverse: Scaling 3d Vision-language Learning For Grounded Scene Understanding, by Baoxiong Jia et al.
SceneVerse: Scaling 3D Vision-Language Learning for Grounded Scene Understanding
by Baoxiong Jia, Yixin Chen, Huangyue Yu, Yan Wang, Xuesong Niu, Tengyu Liu, Qing Li, Siyuan Huang
First submitted to arxiv on: 17 Jan 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Robotics (cs.RO)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper addresses the challenges of grounding language in three-dimensional (3D) scenes, which is a crucial step in developing embodied agents. The authors focus on indoor environments and introduce the SceneVerse dataset, a million-scale collection of 3D vision-language pairs. They also propose the Grounded Pre-training for Scenes (GPS) framework, which allows for unified pre-training for 3D vision-language learning. The authors demonstrate the effectiveness of GPS by achieving state-of-the-art performance on existing 3D visual grounding benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper makes it possible for computers to understand language in 3D environments, like homes or offices. Right now, this is a big challenge because there isn’t much data available that shows how words relate to the 3D world. The authors create a huge dataset called SceneVerse that has about 68,000 3D indoor scenes and 2.5 million pairs of images and text descriptions. They also make a new way to learn from this data called Grounded Pre-training for Scenes (GPS). This helps computers get better at understanding language in 3D environments. |
Keywords
* Artificial intelligence * Grounding
