Matthias Niessner (@MattNiessner) / X

Matthias Niessner

1,787 posts

Matthias Niessner

@MattNiessner

Professor for Visual Computing & Artificial Intelligence @TU_Muenchen Co-Founder @synthesiaIO Co-Founder @SpAItial_AI

Munich, Bavaria

niessnerlab.org/publications.h…

Joined March 2015

Matthias Niessner
@MattNiessner
Nov 28, 2023
(1/2) Check out 𝐌𝐞𝐬𝐡𝐆𝐏𝐓! MeshGPT generates triangle meshes by autoregressively sampling from a transformer model that produces tokens from a learned geometric vocabulary. As a result, we obtain clean and compact meshes :) nihalsid.github.io/mesh-gpt/ youtu.be/UV90O1_69_o
00:00
396K
Matthias Niessner
@MattNiessner
Sep 23, 2021
(1/n) How to start a deep learning project? We use a remarkably streamlined step-by-step process to set up deep learning projects. At the same time, people who are new to deep learning tend to always make the same (avoidable) mistakes. Check out the thread below! 🧵
Matthias Niessner
@MattNiessner
Jul 8, 2023
If you want to do high-impact research, start by reading papers from 20+ years ago. Then, apply these fundamental concepts to modern science. Most importantly, you'll get an advantage over everyone else who is only superficially parsing the latest paper feed on social media.
185K
Matthias Niessner
@MattNiessner
Jan 12, 2021
My default interview question is "how to solve a linear system A x = b". Sadly, the default answer is "gradient descent" w/o further explanation. For a career in AI/ML, knowing linear algebra is a must!
Matthias Niessner
@MattNiessner
May 21, 2022
Newton's method basically solves everything - the Babylonians had already known this for square root computation! Also known as a common tech interview question :)
Matthias Niessner
@MattNiessner
Mar 2, 2021
(1/n) In the past 2.5 years, I received about 1,000 PhD applications. I wanted to share some thoughts, which might be helpful to get into the right program. Experience is from a European perspective but should apply elsewhere. Here's the lessons learned: 👇
Matthias Niessner
@MattNiessner
Dec 20, 2023
Google's Gemini paper has about 950 co-authors :) That's 10 full paper pages that list the author names. In comparison, the method description is only a single page (3 pages if training is included). Don't use "𝑒𝑡 𝑎𝑙." when citing the paper! arxiv.org/abs/2312.11805
372K
Matthias Niessner
@MattNiessner
Sep 3, 2021
PhD student salaries in Germany are quite competitive: standard positions in computer science are about 55k Euro / year (TVL-E13 level 2, 100%, incl. benefits). That's above the entry-level industry average for MA degrees in CS! -> come to Germany, do a PhD - it's worth it :)
Matthias Niessner
@MattNiessner
Dec 4, 2023
(1/2) Check out 𝐆𝐚𝐮𝐬𝐬𝐢𝐚𝐧𝐀𝐯𝐚𝐭𝐚𝐫𝐬: Photorealistic Head Avatars with Rigged 3D Gaussians! We create photorealistic head avatars by animating 3D Gaussians on a parametric face model - edited and rendered in *real-time*! youtu.be/lVEY78RwU_I shenhanqian.github.io/gaussian-avata…
00:00
127K
Matthias Niessner
@MattNiessner
Jun 25, 2023
I'm kind of frustrated to see pseudo AI influencer accounts consistently posting factually wrong information about research. This work has nothing to do with Stable Diffusion or any other diffusion method. It fits a 3D GAN to an image as described in the PanoHead CVPR paper.
Drake Facts
@NewsIn6ix
Jun 23, 2023
Stable Diffusion is now capable of creating photo realistic full 3D models from single images. The amount of ways it could be used in video games and the metaverse blows my mind. AI is getting closer and closer to futuristic sci-fi movies!
00:00
133K
Matthias Niessner
@MattNiessner
Jun 22, 2019
Main lessons learned from #CVPR: - CVPR'17: full autonomous driving is 5 years away - CVPR'18: full autonomous driving is 10 years away - CVPR'19: not so sure when it will happen Obviously, just hearsay :)
Matthias Niessner
@MattNiessner
Dec 20, 2022
PhD studies are highly non-linear in terms of outcome: students often work on projects for long periods without any concrete outcome until the right idea hits. What really stands out though are the PhDs who never give up, relentlessly trying new ideas, until something sticks!
95K
Matthias Niessner
@MattNiessner
May 2, 2025
📢Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction📢 -> highly accurate face reconstruction by training powerful VITs via surface normals and UV-coordinates estimation. The geometric cues from our 2D foundation model backbone constrain the 3DMM
00:00
62K
Matthias Niessner
@MattNiessner
Sep 5, 2021
There's still this illusion that "AI" is somewhat intelligent: these are models fitted to large datasets using non-linear optimizations, hoping to replicate their distributions. While these are incredibly useful tools, there is nothing we would consider human-like intelligence.