Zhiwen(Aaron) Fan (@zhiwen_fan_)

2024-04-02 | โค๏ธ 80 | ๐Ÿ” 15


  • Dive into our latest research on Multimodal SLAM: unposed camera images + inertial measurements โ†’ real-time rendering + dense 3D map.
  • The new UT-MM dataset, a multi-modal one from a mobile robot + a camera + an inertial measurement unit.

๐Ÿ‘‰Paper: https://arxiv.org/abs/2404.00923 ๐Ÿ‘‰Project: https://vita-group.github.io/MM3DGS-SLAM/

๋ฏธ๋””์–ด

video


์š”์•ฝ

๋น„์ •ํ•ฉ(unposed) ์นด๋ฉ”๋ผ ์ด๋ฏธ์ง€์™€ IMU(๊ด€์„ฑ) ์ธก์ •์„ ๊ฒฐํ•ฉํ•ด ์‹ค์‹œ๊ฐ„ ๋ Œ๋”๋ง๊ณผ ์กฐ๋ฐ€ํ•œ 3D ๋งต ์ƒ์„ฑ์„ ์ˆ˜ํ–‰ํ•˜๋Š” ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ SLAM ์—ฐ๊ตฌ๋ฅผ ์†Œ๊ฐœํ•œ๋‹ค. ์ด๋ฏธ์ง€์—์„œ๋„ RGB/Depth ์ž…๋ ฅ๊ณผ ํ•จ๊ป˜ 3D ํฌ์ธํŠธ ๋ฐ ๊ถค์ (Tracked vs Ground Truth) ์‹œ๊ฐํ™”๋ฅผ ํ†ตํ•ด ์ถ”์ ยท์žฌ๊ตฌ์„ฑ ์„ฑ๋Šฅ์„ ๋ณด์—ฌ์ค€๋‹ค.

Tags

domain-vision-3d domain-robotics domain-vlm domain-xr domain-rendering