Matthias Niessner @MattNiessner Twitter profile

Last Seen Profiles

@AzraMansoorAhm1

@ggazza74

@karinaGsalazar

@Medusaa_Jr

@Parlons_PP

@SHPAthletics

@WIS_Tobias

@Eve_Goddess_

@aidounaiwu

@Everdeen_060786

@Shahzad6068

@H__a_p_py

@polarbeeru

@StRitaOCSB

@MathDemarcy

@ScottFieldsMT

@SlwmyW

@ittsdestany

@AHoensbroech

@Noemie_Bilas

@Itsjustwitch

@BasicAshHoe

@NoGoD_Official

@La_Felguera

@Mh32843005H

@tmg_just

@norasea_

@NickAndo1004

@RealStamkos91

@lisautism

@NavinNooli

@pem_peem8

@_neededsaving

@evghenia_v

@CamilleGrenu

@scheist_

Matthias Niessner

@MattNiessner

3 years

(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! 🧵

26

538

2K

Matthias Niessner

@MattNiessner

6 months

(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 :)

24

427

2K

Matthias Niessner

@MattNiessner

10 months

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.

33

201

2K

Matthias Niessner

@MattNiessner

3 years

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!

72

80

1K

Matthias Niessner

@MattNiessner

2 years

Newton's method basically solves everything - the Babylonians had already known this for square root computation! Also known as a common tech interview question :)

14

170

1K

Matthias Niessner

@MattNiessner

3 years

(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: 👇

31

319

1K

Matthias Niessner

@MattNiessner

5 months

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!

43

131

1K

Matthias Niessner

@MattNiessner

5 months

(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*!

15

207

932

Matthias Niessner

@MattNiessner

3 years

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 :)

43

83

912

Matthias Niessner

@MattNiessner

11 months

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.

AI Daily

@AIDailyNewsNow

11 months

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!

31

161

669

27

85

848

Matthias Niessner

@MattNiessner

5 years

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 :)

8

181

845

Matthias Niessner

@MattNiessner

1 year

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!

17

91

826

Matthias Niessner

@MattNiessner

3 years

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.

53

103

771

Matthias Niessner

@MattNiessner

10 months

Can we apply diffusion directly on MLP weights? Yes!!! We show generalization to new 3D shapes and 4D mesh animations: "HyperDiffusion: Generating Implicit Neural Fields with Weight-Space Diffusion" #ICCV2023 ! Project: Video:

8

149

631

Matthias Niessner

@MattNiessner

2 months

Our group has fourteen papers accepted at #CVPR '2024! Exciting topics: lots of diffusion & transformers focusing on generative AI for image synthesis, geometry generation, and many more - check it out! I'm so proud of everyone involved - let's go 🚀🚀

8

61

624

Matthias Niessner

@MattNiessner

3 years

The "Facebook -> Meta" rebranding is quite a big deal for computer vision & graphics research. It's a long-term commitment by a trillion dollar tech company to virtual 3D cyberspaces (aka metaverse), driven by virtual/augmented reality tech, which is great for our field!

16

65

590

Matthias Niessner

@MattNiessner

3 years

The main idea of Machine Learning is that everything is simply learned from data instead of tedious handcrafting features. The irony is that we are now doing more empirical research than ever, spending most of our time with data engineering and hyper parameter tuning.

11

76

586

Matthias Niessner

@MattNiessner

5 months

(1/2) 📢📢𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧𝐀𝐯𝐚𝐭𝐚𝐫𝐬 📢📢 High-fidelity 3D head avatars with precise control over viewpoint, expression, and pose. -> Our parametric 3D model enables control & consistency + 2D diffusion makes it photoreal.

7

125

584

Matthias Niessner

@MattNiessner

5 months

(1/2) Check out "𝐏𝐨𝐥𝐲𝐃𝐢𝐟𝐟: Generating 3D Polygonal Meshes with Diffusion Models"! Our model operates directly on the polygons of 3D meshes and generates novel shapes as output through an iterative diffusion process.

5

109

574

Matthias Niessner

@MattNiessner

4 years

Modeling 3D Shapes by Reinforcement Learning! We train agents that aim to mimic human modeling steps by self supervision; agents are rewarded by evaluating the the goodness of the final model :) Video: Paper:

10

134

573

Matthias Niessner

@MattNiessner

8 months

PhD graduates in AI mostly take boring jobs at big tech companies due to short-term monetary incentives. While understandable to some degree, it's also quite sad to see so many great researchers 'disappear' and give up their talent - join or do your own startup instead!

47

46

573

Matthias Niessner

@MattNiessner

1 year

Excited to share @Normanisation 's DiffRF: Rendering-guided 3D Radiance Field Diffusion #CVPR2023 highlight! 2D diffusion is great, but what about 3D? We show radiance field diffusion with rendering guidance for consistent and editable 3D synthesis. Vid:

3

153

570

Matthias Niessner

@MattNiessner

3 years

Common issue in deep learning projects is that complex method designs are adopted before basic debugging. Often this leads to situations where the data loader still loads black images, but we've already tried 10 loss function.... cuz a paper claimed these would improve results.

12

49

563

Matthias Niessner

@MattNiessner

1 year

Most common question from PhD applicants: "How many GPUs do you have in the lab?" I used to prepare answers about good advising, research impact, and career opportunities... but seems priorities have somewhat changed :)

23

18

544

Matthias Niessner

@MattNiessner

2 years

(1/n) Key to successful projects in Deep Learning are fast turnaround times of experiments. For large models, training often takes several days or even weeks, and it might need countless runs to find hyperparameters that yield good results. How to get things fast? A thread🧵

11

94

540

Matthias Niessner

@MattNiessner

5 years

Check out DeepVoxels: Instead of 2D convs, like in most GANs, we lift image features to a latent 3D space. This way, we obtain a 3D representation from which we can generate consistent novel views. Don't learn 3D operations with 2D convs -- use 3D :)

8

138

523

Matthias Niessner

@MattNiessner

11 months

Introducing #ProfGPT , my new AI Avatar by @synthesiaIO ! Getting closer towards fully automating myself :)

13

48

517

Matthias Niessner

@MattNiessner

2 years

Lots of ML theory behind deep learning, but the truth is that what made things work was clever engineering & programming abstractions (Caffe, TensorFlow, PyTorch, ...). These design choices now dictate our workflow, and the next major AI breakthrough will require similar tools.

14

52

513

Matthias Niessner

@MattNiessner

2 years

Twitter seems to have changed it's recommendation algorithm in the recent days. My feed now shows random 'viral' tweets that I have zero interest in - similar to what Facebook has been done a while ago, which made it practically unusable... Anyone else with the same issue?

35

14

483

Matthias Niessner

@MattNiessner

1 year

Publishing one great paper requires significantly less effort than publishing ten mediocre papers, yet only the great paper will be remembered. In other words, focusing on a single high-impact work definitely pays off as it yields disproportionately high returns on your efforts.

20

38

477

Matthias Niessner

@MattNiessner

2 years

(1/n) We often hear that AI & machine learning produce great results but we don't understanding why. Specifically, many consider neural networks to be black boxes that no one understands. However, I don't think that's true; in fact, research has quite some insights. A thread 🧵

11

74

467

Matthias Niessner

@MattNiessner

2 years

(1/2) How to store 3D scenes/grids efficiently? Just hash occupied voxels, 𝐻(𝑥,𝑦,𝑧)=(𝑥⋅𝑝_1⨁ 𝑦⋅𝑝_2⨁ 𝑧⋅𝑝_3) mod 𝑛, and store them in a linear array! The voxel hashing method proposed this for 3D recon in 2013, but it's used in many apps such as sparse conv nets!

7

59

459

Matthias Niessner

@MattNiessner

10 months

So far, every graphics researcher I know has been able to run a neural network. On the other hand, ask a 'Generative AI' researcher to explain the rendering equation...

23

39

458

Matthias Niessner

@MattNiessner

6 months

Writing a good paper intro is difficult. I mostly recommend a 4-paragraph intro: 1) Motivation: Task description / why is it important? 2) Challenge: Why is problem so difficult? 3) Trends: How does SotA approach it? What's missing? 4) Method: How do you solve it? Contributions!

4

60

459

Matthias Niessner

@MattNiessner

2 years

Check out AutoRF: Learning 3D Object Radiance Fields from Single View Observations #CVPR2022 by @Normanisation . Key idea: we learn shape & appearance priors from single-view training samples by exploiting machine-annotated labels!

1

76

419

Matthias Niessner

@MattNiessner

5 months

(1/3) 𝐕𝐨𝐱𝐞𝐥 𝐇𝐚𝐬𝐡𝐢𝐧𝐠 received the Test-of-Time Award @SIGGRAPHAsia ! What an honor together with @MZollhoefer @izadi_shahram @mcstammi Voxel hashing is a sparse & efficient data structure for 3D scenes/grids! What's the core idea and why is still relevant today? ⬇️

11

48

434

Matthias Niessner

@MattNiessner

11 months

We have several PhD and PostDoc openings in our Visual Computing & AI Lab at TU Munich! Topics focus on Generative AI for Vision/Graphics, including NeRFs, Neural Scene Representations, Diffusion Models, LLMs, etc. Please share & apply by July 31st :)

4

119

420

Matthias Niessner

@MattNiessner

2 years

Check out Texturify at #ECCV2022 ! Our graph generative model produces high-quality textures on meshes, trained entirely unsupervised! For training, we only need ShapeNet (no textures) and a set of real images. The texture model is then learned through differentiable rendering.

4

67

420

Matthias Niessner

@MattNiessner

1 year

(1/2) 📢📢📢NeRSemble #SIGGRAPH '23! We reconstruct dynamic radiance fields for high-quality novel view synthesis of human heads. Key is a deformation field and an ensemble of multi-resolution hash encodings to model coarse & fine-scale deformations.

3

102

404

Matthias Niessner

@MattNiessner

10 months

Our "Introduction to Deep Learning" (I2DL) 2023 lecture is now on YouTube! We cover DL fundamentals - backprop, SGD, losses, CNNs, etc. - as well as advanced concepts such as RNNs/LSTMs transformers, generative models (GANs, diffusion)! Check it out:

Introduction to Deep Learning (I2DL 2023) - 1. Introduction

Website & Slides: https://niessner.github.io/I2DL/Introduction to Deep Learning (I2DL) - Lecture 1TUM Summer Semester 2023Prof. Niessner

www.youtube.com

4

102

404

Matthias Niessner

@MattNiessner

8 months

Can we match visual features jointly across multiple frames? Yes! @barbara_roessle 's #ICCV2023 paper proposes a differentiable pose optimization for end2end feature matching across multiple frames, thus obtaining better poses!

1

92

393

Matthias Niessner

@MattNiessner

5 months

(1/2) Intrinsic Image Diffusion for Single-view Material Estimation! We propose a probabilistic diffusion model to handle material & lighting ambiguities. We obtain sharp material estimates and facilitate high-fidelity relighting.

1

77

390

Matthias Niessner

@MattNiessner

7 months

Diffusion models are awesome! Check out our survey on 𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬 𝐟𝐨𝐫 𝐕𝐢𝐬𝐮𝐚𝐥 𝐂𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠! We give an introduction to diffusion models and highlight how they are used by state-of-the-art methods in graphics and vision.

5

91

386

Matthias Niessner

@MattNiessner

3 years

How to solve a linear system? High-school student applying for university admission: "Gaussian Elimination" Master student applying for PhD positions after taking ML lecture: "Gradient Descent" Apparently students are un-learning optimization... *sigh*

18

28

373

Matthias Niessner

@MattNiessner

2 months

(1/3) Can we turn text-to-image models into photorealistic 3D generators? ViewDiff ( #CVPR2024 ) produces realistic, multi-view consistent images of real-world 3D objects in authentic surroundings. Website Video How does it work?

4

85

377

Matthias Niessner

@MattNiessner

3 years

Often students become less communicative when they struggle to get a project idea working. However, reaching out to advisors is particularly critical in these situations in order to find solutions. That's much more important than giving updates when everything is great and fine.

3

22

370

Matthias Niessner

@MattNiessner

10 months

(1/4) Excited to share our #ICCV2023 paper Text2Room! We generate scene-scale textured 3D meshes from a given text prompt leveraging 2D text-to-image models such as StableDiffusion. Project: Code: Video:

4

89

369

Matthias Niessner

@MattNiessner

6 months

We have new PhD & PostDoc openings in our Visual Computing & AI Lab at TU Munich! Topics have a strong focus on Generative AI for Vision/Graphics, including NeRFs, Neural Scene Representations, Diffusion Models, LLMs, etc. Please share & apply :)

4

82

367

Matthias Niessner

@MattNiessner

4 months

We are looking for interns & visiting researchers in our Visual Computing & AI Lab at TU Munich! Topics focus on Generative AI, NeRFs, Neural Scene Representations, Diffusion Models, LLMs, etc. Please share & apply by Feb 29th :) Details: - We focus on…

3

56

353

Matthias Niessner

@MattNiessner

3 years

We have a record enrollment of over 1500 students in our "Introduction to Deep Learning" (I2DL) lecture. Sadly, the lecture is online; hopefully the last time... On the bright side, a lot of lecture content is publicly available!

Introduction to Deep Learning (I2DL 2020) - Technical University of Munich - Prof. Niessner, Prof....

www.youtube.com

2

60

337

Matthias Niessner

@MattNiessner

1 year

(1/n) How to obtain a photorealistic digital replica of an environment? Well, run a 3D reconstruction and use it for re-rendering. Neural rendering can fix imperfections during rendering. Neural Radiance Fields (NeRF) was a game changer! How and what's missing? A thread🧵

2

44

322

Matthias Niessner

@MattNiessner

3 years

In order to solve computer vision, one needs to understand computer graphics. "What I cannot create, I do not understand" -> make sure to take graphics courses!

14

26

321

Matthias Niessner

@MattNiessner

2 years

Main failure cases in research: - spending too much effort on a bad idea - giving up too quickly on a good idea

14

27

311

Matthias Niessner

@MattNiessner

3 years

Check out @BozicAljaz work TransformerFusion, an online monocular RGB scene reconstruction using a transformer to learn to attend to the most relevant pixel observations. Project page: Paper: Video:

4

60

311

Matthias Niessner

@MattNiessner

3 months

(1/n) 𝐒𝐨𝐫𝐚 generates stunning videos and is a game changer! - But how does it work technically? - What’s different from existing video diffusion? - How did we get there? Lots of speculation - here's my take from a technical perspective! 🧵

3

65

305

Matthias Niessner

@MattNiessner

2 years

Life is back at the University ❤️ Thanks to all students for making it such a great place!

11

2

303

Matthias Niessner

@MattNiessner

5 months

(1/2) 📢📢𝗦𝗰𝗲𝗻𝗲𝗧𝗲𝘅 📢📢 Given scene geometry and text prompt -> SceneTex generates high-quality textures. Main idea: directly optimize scene texture with gradients from a score-distillation objective with view sampling.

4

76

297

Matthias Niessner

@MattNiessner

4 years

All videos and slides for our Introduction to Deep Learning Lecture (I2DL) with @lealtaixe are now public! Covering: ML basics, Neural Networks, Optimization, Training Processes, RNNs.. Videos: Slides: #deeplearning @TU_Muenchen

Introduction to Deep Learning - 1. Introduction (Summer 2020)

Website: https://niessner.github.io/I2DL/Slides: https://niessner.github.io/I2DL/slides/1.Intro.pdfIntroduction to Deep Learning (I2DL) - Lecture 1TUM Summer...

www.youtube.com

3

102

292

Matthias Niessner

@MattNiessner

1 year

Our GPU compute cluster got a little upgrade: 20x brand-new A100 - 80GB variants! Looking forward to training larger models :)

7

6

280

Matthias Niessner

@MattNiessner

3 years

The main reason why tech companies pay insane salaries and VCs invest crazy $$$ in AI/ML is simple: it's all about scalability! Do it once, sell to billions! For large platforms, this means that even small improvements make a huge difference and justify pretty much any cost.

7

32

278

Matthias Niessner

@MattNiessner

2 years

Excited to share @DejanAzinovic 's Neural RGB-D Surface Reconstruction #CVPR2022 ! We reformulate NeRF to operate on a signed distance field to obtain high-quality 3D surface reconstructions. @rmbrualla @DanBGoldman @JustusThies

3

62

278

Matthias Niessner

@MattNiessner

1 year

Senior PhD advisor, who didn't know about the student's paper submission, when the paper gets accepted.

1

15

275

Matthias Niessner

@MattNiessner

11 months

Check out my Volumetric Capture! The avatar was recorded in the studio of the AIT Lab in Zurich - thx @OHilliges . Their capture system is located in a green room and records body movements with more than 100 spherically arranged high-speed RGB cameras.

5

28

274

Matthias Niessner

@MattNiessner

1 year

If you want to become famous in academia, you can't do what everyone else does. Instead, you have to make everyone else work on your research problems that you define. The reason why this is so challenging is that we researchers tend to mimic successful research from others.

8

32

266

Matthias Niessner

@MattNiessner

5 years

The cab driver told us there are two major events in Seoul: BTS concert and ICCV. Vision folks we made it! We're at the same level than the mainstream kpop band - Nerds, we're cool now :) #ICCV #iccv2019

0

33

263

Matthias Niessner

@MattNiessner

5 months

(1/2) Check out DPHMs: Diffusion Parametric Head Models for Depth-based Tracking! 3D head reconstruction from noisy & sparse depth? Our diffusion prior constrains identity/expression latents, and maps them to high-quality samples.

1

61

259

Matthias Niessner

@MattNiessner

8 months

(1/2) How to use GANs for high-quality NeRF reconstruction? GANeRF proposes an adversarial rendering formulation whose gradients constrain a 3D NeRF optimization: highly-realistic & view-consistent renderings! #SIGGRAPHAsia2023

3

54

256

Matthias Niessner

@MattNiessner

3 years

No matter how brilliant or famous a researcher, doing good research requires continuously reading up on the latest literature (in particularly in fast moving fields). I can only recommend to spend a dedicated amount of time per week reading papers.

4

17

254

Matthias Niessner

@MattNiessner

10 months

We released our ActorsHQ dataset from our #SIGGRAPH work: "HumanRF: High-Fidelity Neural Radiance Fields for Humans in Motion". ActorsHQ provides high-fidelity multi-view captures at 12MP resolution from 160 cameras with per-frame mesh reconstructions.

2

57

253

Matthias Niessner

@MattNiessner

3 years

7 Papers provisionally accepted to #CVPR2021 from our lab! Great job everyone!!! (assuming I typed in the IDs correctly...)

3

2

254

Matthias Niessner

@MattNiessner

2 years

Very happy that six paper got accepted at #CVPR '2022! Also great news: this will be the first conference in a while where we'll physically attend and catch up with so many friends - see you all in New Orleans!!! Congrats to all the students and collaborators :)

5

13

249

Matthias Niessner

@MattNiessner

2 years

When finding new research projects, we all make the same mistake: we read the most popular and most recent papers, and then directly build on them. Often this leads to incremental works in overcrowded areas, having to compete for minor improvements with many other researchers.

6

29

245

Matthias Niessner

@MattNiessner

3 years

- Research is much more a team sport than the ingenuity of an individual - Research relies on collaboration within one and/or different projects - Research by definition is incremental as it builds on existing work - Research is hard work / persistence over long time periods

3

29

249

Matthias Niessner

@MattNiessner

1 year

📢Panoptic Lifting for 3D Scene Understanding with Neural Fields #CVPR23 highlight! Given only posed RGB images of a scene, we optimize a panoptic radiance field representing color, depth, semantics, and instances at any point in space. Vid: @yawarnihal

2

64

248

Matthias Niessner

@MattNiessner

1 year

Research projects have four steps: 1) idea finding 2) execution 3) writeup 4) presentation What's important: no matter of 2) 3) 4), the idea defines the maximum impact of a project. During a project you can practically only scale back. So when thinking about an idea: aim high!

6

17

247

Matthias Niessner

@MattNiessner

11 months

(1/2) We released our Neural Parametric Head Models (NPHM) dataset from our #CVPR2023 paper! It includes over 5600 high-fidelity 3D scans of human heads from 272 subjects - all publicly available! Check it out!

5

56

241

Matthias Niessner

@MattNiessner

5 years

Coolest #Siggraph2019 Paper - Symmetric ICP to align two Point Clouds! Just like Point-to-Plane ICP, except uses both normals: min[(p-q)*(n_p+n_q)]^2 Just one line of code, so simple, works better, just cool :) By Szymon Rusinkiewicz:

3

59

244

Matthias Niessner

@MattNiessner

1 year

Very excited to share the seven #CVPR '23 papers from our lab! Super cool topics, including 3D diffusion models, NeRFs, faces, and 3D scenes! Will post details about each project soon - check out the projects: Congrats to all students & collaborators :)

5

45

240

Matthias Niessner

@MattNiessner

1 year

(1/2) Excited to share "Learning Neural Parametric Head Models" #CVPR2023 ! We capture over 5200 high-quality 3D human head scans from which we build a neural parametric head model that disentangles & expressions and deformations.

2

72

240

Matthias Niessner

@MattNiessner

5 months

Merry Christmas & Happy Holidays! May your diffusion denoise, your Gaussians splat well, and your GPUs provide enough warm during the cold winter season 🎄🎅☃️

1

8

240

Matthias Niessner

@MattNiessner

2 years

We released the videos & slides of our 3D Scanning & Motion Capture Lecture at TUM! This was really fun to teach, covering fundamental 3D reconstruction and optimization techniques - enjoy watching :) Videos: Slides & Content:

3D Scanning & Motion Capture: 1. 3DConcepts

Website: https://niessner.github.io/3DScanning/

www.youtube.com

2

69

237

Matthias Niessner

@MattNiessner

10 months

(1/2) Happy to announce that Text2Tex has been accepted at #ICCV2023 🎉 Taking a mesh and a text prompt as input, Text2Tex generates high quality textures - it's fully automated and easy to scale to many models! Project: Video:

1

63

239

Matthias Niessner

@MattNiessner

4 months

Check out 𝐌𝐨𝐭𝐢𝐨𝐧𝟐𝐕𝐞𝐜𝐒𝐞𝐭𝐬, a 4D diffusion model for dynamic surface reconstruction from imperfect observations of sparse, noisy, or partial point clouds. Main idea: we represent time-varying shapes via 4D neural representation with latent vector sets, and then…

2

38

238

Matthias Niessner

@MattNiessner

10 months

(1/2) We released the NeRSemble data from our #SIGGRAPH2023 work! We provide a high-fidelity multi-view video dataset with >5800 recordings of faces of 269 different people captured from 16 high-end cameras: - global shutter - 73fps - 7.1MP Try it out:

2

56

237

Matthias Niessner

@MattNiessner

2 years

Kicking off our TUM AI - Lecture Series in 2022 on Monday with no other than Jitendra Malik! He will be talking about "Learning to Walk With Vision and Proprioception". Live stream here: 7pm GMT+1 / 10am PST (Monday Jan 17th)

1

35

237

Matthias Niessner

@MattNiessner

2 years

Our state-of-the-art report on "Advances in Neural Rendering" is being presented at #Eurographics2022 ! Live recording: A great overview about hot topics & open challenges in neural rendering, neural radiance fields, and many more!

4

58

235

Matthias Niessner

@MattNiessner

2 years

Too many successful junior faculty are leaving academia, or are essentially working full-time at companies - a very problematic situation. Universities seem to do very little, leaving young researchers in uncertain career situations, overwhelming them with bureaucracy, etc.

8

17

232

Matthias Niessner

@MattNiessner

10 months

Super happy that our group has eight papers accepted at #ICCV '2023! Fantastic topics: 3D generative models, diffusion, NeRFs, neural rendering, and many more - check it out! Very excited about all works! Congrats to all students and collaborators :)

5

28

234

Matthias Niessner

@MattNiessner

1 year

Debugging research code before the paper deadline.

2

17

230

Matthias Niessner

@MattNiessner

5 years

Tenured and promoted to a W3-Professor with my own Chair of "Visual Computing & AI" at TUM! I finally decided to take the TUM offer, which was too good to refuse. We will continue building our amazing lab in cutting-edge research at the intersection of vision, graphics, and ML.

16

7

227

Matthias Niessner

@MattNiessner

3 years

We released the code for "SceneFormer: Indoor Scene Generation with Transformers" by @XinpengWang_ and @chandan__yes From a room layout or text input, our transformer model generates realistic indoor scenes! Code: Project:

2

52

226

Matthias Niessner

@MattNiessner

3 years

During your PhD, doing an internship can be a fantastic experience (I really enjoyed my own experiences). But it is critical to make ensure that - you can do open research & publish the results - the research is aligned with your PhD goals

2

8

226

Matthias Niessner

@MattNiessner

3 years

We made a website of our TUM AI Lecture series: - Videos of all previous speakers online! - Schedule with upcoming speakers! Talks are always streamed live on Friday's (see links above) - 9am PST / 6pm CET. Really cool talks - must watch -:)

1

48

222

Matthias Niessner

@MattNiessner

11 months

Super excited - our company @synthesiaIO is now a unicorn🦄🦄🦄 !!! We are making important steps in advancing cutting-edge research in generative AI. Check out our first publications this year at #CVPR and #SIGGRAPH -- a lot of cool stuff is in the pipeline and yet to come!

Synthesia 🎥

@synthesiaIO

11 months

Synthesia just raised $90 million in Series C funding led by @Accel and with that… officially joined the unicorn club 🦄 🦄 🦄 Back in 2017, no one really understood our vision.. When @vriparbelli and @Stjerrild first used the term "generative video," people just didn't get…

30

111

431

11

17

223

Matthias Niessner

@MattNiessner

1 year

Excited to share our new work "Learning 3D Scene Priors with 2D Supervision" #CVPR2023 ! 3D labels are costly! We learn priors of object semantics and shapes in 3D scenes with only 2D supervision. @yinyu_nie @angelaqdai @XiaogHan

0

52

219

Matthias Niessner

@MattNiessner

5 months

(1/2) 📢📢MonoNPHM 📢📢 Our neural parametric head model disentangles geometry, appearance, and expressions. -> At test time, we can fit our model to images or videos, and obtain high-fidelity dynamic 3D head reconstructions.

3

57

219

Matthias Niessner

@MattNiessner

3 years

How to successfully publish a paper at a top-tier venue like @cvpr , @siggraph , @NeurIPSConf ? Key is to understand the psychology of your research area: what sells? what baselines are expected? This making it particularly hard for new researchers even with a great idea/method.

6

22

218

Matthias Niessner

@MattNiessner

1 year

Preparing our "Introduction to Deep Learning" course! Almost done automating myself with my #AI avatar :) Avatar by @synthesiaIO #AI #GenerativeAI

3

11

215

Matthias Niessner

@MattNiessner

2 years

Dancefloor #CVPR '2018 vs #ECCV '2022: computer vision has evolved - are we cool now?

7

16

213

Matthias Niessner

@MattNiessner

2 years

Super happy to have seven papers accepted at #ECCV '2022! Will post more updates soon - very excited about all the work, and looking forward to seeing you all in Tel-Aviv!!! Congrats to all students and collaborators :)

7

22

212