robocrunch
Frank Dellaert
@fdellaert
Professor at Georgia Tech: Robotics & Computer Vision. Part-time research scientist in Google AI. Before: sabbaticals at KUL, Skydio, Facebook B*8.
Tweets by Frank Dellaert
Our #CVPR22 paper on Panoptic NeRF is now released on arxiv. TL;DR: view synthesis + semantics on “stuff” and objects in the scene. Object-based NeRFs also allow for editing/moving/removing.
Shared by
Frank Dellaert
at
5/10/2022
Fully agreed; in addition, model your audience as having ADD and a desperate need to be entertained :-) I always thought Princeton’s (ex-)Prof X.’s students, like @SongShuran , did an amazing job at this.
Shared by
Frank Dellaert
at
4/20/2022
Another paper I missed (in session 10): NeRD (https://t.co/KdK2mGEByU) or “Neural Reflectance Decomposition” uses physically-based rendering to decompose the scene into spatially varying BRDF material properties, enabling re-lighting of the scene. (N+9) #ICCV2021
Shared by
Frank Dellaert
at
10/12/2021
A Conditional NeRF paper drop from UCL, incl. @LourdesAgapito . This is getting more and more popular. What other conditional NeRF variants do y’all know of?
Shared by
Frank Dellaert
at
9/7/2021
Lie groups are actually not needed, turns out you just need manifolds! @gtsam4 was based on Lie groups for a few years, but then we found manifolds to optimize over (3D planes, directions in 3D, essential matrix) that do *not* have group structure. Check out great book by Absil++
Shared by
Frank Dellaert
at
9/6/2021
Very nice work on multi-view surface reconstruction using deep SDF representations. Results on the challenging Tanks and Temples dataset…
Shared by
Frank Dellaert
at
8/24/2021
And another effort in which deep learning speeds up optimization, this time mixed integer programming…
Shared by
Frank Dellaert
at
7/26/2021
Robotics and arts workshop is now over, but you can see the exhibits here as a permanent web gallery: https://t.co/yPLf3lY2SP
Shared by
Frank Dellaert
at
7/14/2021
Our RSS 2021 Workshop on Robotics x Arts is *next week*. Gathertown exhibits and Bluejeans panel. https://t.co/Z4gTGx6HNb
Shared by
Frank Dellaert
at
7/7/2021
I actually meant the fact that when coding in pytorch or tensorflow you always have to account for the batch dimensions. It’s like programming in assembly all over again. We should have something #halide - like that abstracts all that away. Does that exist??
Shared by
Frank Dellaert
at
6/19/2021
My Annual Reviews article on Factor Graphs in Robotics is finally out with a publicly accessible link: https://t.co/Xc0RMXyeYY
Shared by
Frank Dellaert
at
5/27/2021