Chris Offner @chrisoffner3d Twitter profile

Pinned Tweet

Chris Offner

2 months

In real-world robotics or autonomous vehicle use, vibrations, shocks, temperature changes, etc. can misalign the relative position of stereo cameras, so I'm looking into online stereo extrinsics self-calibration. Via Blender and Python I can get pixel-perfect keypoint matches.

3

9

94

Last Seen Profiles

@EmojiMashupBot

@kaydaustin

@haru_oh

@bjan9

@Kylewalkersongs

@MJamille

@gulcanozenkucuk

@PayhtonPorter

@azamsharp

@cynorrhodons

@illshin

@tsurumachi_ren

@leleketpis

@Ramunepc

@harucolacola

@fan_hongda

@maloneathletics

@_Sehaj_3

@_Yare_Yare_Daze

@kystackzzz

@_Swaggy0

@MarkGrote

@ahmontewatkins

@aatishparashar

@_Jhinxx

@Vanreito

@Yuki_cZ7Sl288

@ZTXofficial

@Quintaine

@_vihkie_

@_LightRoomLDN

@_Ejjenkins

@_donte__

@Mwai999

@BdrAlra

@srhdoost

Chris Offner

@chrisoffner3d

20 days

Someone turned a bunch of iconic movie scenes into Gaussian splats. I've been thinking for a while which iconic movie shots would lend themselves best to this, i.e. long and stable tracking shots through scenes with little to no movement.

12

107

641

Chris Offner

@chrisoffner3d

6 months

Playing with self-attention in latent diffusion models. This animation illustrates that the model learns to represent 3d scene properties like depth and object semantics. The red dot shows which pixel's self-attention map we're seeing. #Diffusion #DeepLearning #StableDiffusion

13

26

284

Chris Offner

@chrisoffner3d

7 months

Yesterday I got to present the #GaussianSplatting paper by @Snosixtytwo , @GKopanas and their co-authors to our seminar on Advanced Topics in Vision. The dandelion on made for a lovely title slide.

6

9

135

Chris Offner

@chrisoffner3d

1 month

I'm learning about normalising flows and flow matching and found these fantastic animations on . Not sure who created them, possibly the FFJORD authors ()? Great visualisations!

2

13

117

Chris Offner

@chrisoffner3d

28 days

You vs. the disparity map she told you not to worry about

1

6

94

Chris Offner

@chrisoffner3d

14 days

"mrcal is a generic toolkit built to solve the calibration and SFM-like problems we encounter at NASA/JPL" Great documentation, tutorial, and tools for camera calibration:

1

9

87

Chris Offner

@chrisoffner3d

2 months

3D Gaussian Splat of ETH Zurich's main building.

4

3

74

Chris Offner

@chrisoffner3d

11 days

I have a sadistic urge to test how various feature matchers and stereo matching algorithms perform on scenes like this. 😅

9

4

65

Chris Offner

@chrisoffner3d

1 month

@fchollet “A good science fiction story should be able to predict not the automobile but the traffic jam.” – Frederik Pohl

Chris Offner

@chrisoffner3d

1 month

"A.I.-Generated Garbage Is Polluting Our Culture" Will we act against the pollution of our information environment as halfheartedly and ineffectively as we've acted against the pollution of our natural environment?

0

3

13

2

1

63

Chris Offner

@chrisoffner3d

1 month

On the left we see the disparity maps from the initial calibration. However, the stereo rig has since been de-calibrated through shocks and vibration during use. When trying to perform per-frame camera extrinsics re-estimation, I get more complete per-frame disparity maps but…

1

3

45

Chris Offner

@chrisoffner3d

1 month

Went for a hike yesterday and came across this pretty little Tell's Chapel by the lakeside. Chucked five photos into #DUSt3R . Looks like the benches would require more.

2

3

39

Chris Offner

@chrisoffner3d

2 months

@riverakid1 @JeromeRevaud @bchidlovskii @Vinc3nt_Leroy Witchcraft. 😄 I just ran it on my own images (after changing all mentions of CUDA to [device = "mps" if () else "cpu"]) and wow, this is super cool! 🤩

1

6

32

Chris Offner

@chrisoffner3d

2 months

Every student with only a MacBook to work on will love you for adding an MPS check to your PyTorch device assignment. 🙃

0

4

31

Chris Offner

@chrisoffner3d

27 days

Exchange with my girlfriend: "I hate it when DL-based methods that are essentially magic work better than well-understood algorithms in 3D vision." - "Well then you're likely going to hate the next decades of your life." 🫠

4

0

29

Chris Offner

@chrisoffner3d

2 months

Awesome looking course by @vincesitzmann . Thank you for making this publicly available! :) (Don't get discouraged by the "broken" first recording, the subsequent ones work fine)

0

3

18

Chris Offner

@chrisoffner3d

8 months

@yuliangxiu @ETH The rumours around this tragedy are deeply unsettling and I hope the circumstances of her passing are investigated with great urgency and diligence.

1

18

Chris Offner

@chrisoffner3d

5 months

@zdeborova I'd choose ML explained on a blackboard over ML slides every time!

0

18

Chris Offner

@chrisoffner3d

2 months

@JeromeRevaud I thought you must be joking so I ran it on these two of my own pictures (excuse my dirty dishes 😉) and wow, it works beautifully.

1

4

16

Chris Offner

@chrisoffner3d

29 days

@ykilcher Exactly what @DSMarkovits ’ book is about.

1

0

15

Chris Offner

@chrisoffner3d

1 month

"A.I.-Generated Garbage Is Polluting Our Culture" Will we act against the pollution of our information environment as halfheartedly and ineffectively as we've acted against the pollution of our natural environment?

0

3

13

Chris Offner

@chrisoffner3d

7 months

The scan I did of Einstein's locker here at #ETHZurich can now be explored interactively on #GaussianSplatting #NeRF #LumaAI

1

2

13

Chris Offner

@chrisoffner3d

7 months

@RadianceFields @Snosixtytwo @GKopanas It wasn't recorded yesterday but I might record it myself and put it online some time in the coming weeks. It's only a minor additional effort for me but might serve many more people than just the couple dozen that took the in-person seminar.

1

0

12

Chris Offner

@chrisoffner3d

2 months

@JeromeRevaud 😳

1

11

Chris Offner

@chrisoffner3d

5 months

This looks like a wonderful complement to OpenAI’s “Spinning Up” tutorial introduction to deep reinforcement learning, with a somewhat different but equally interesting structure.

Autonomous Vision Group

@AutoVisionGroup

5 months

Despite its relevance, it can be hard to get into RL. This motivated us to write "An Invitation to Deep Reinforcement Learning", a tutorial for readers with only basic ML knowledge. Huge kudos to @bern_jaeger for this effort. Help us to spread the word!

4

81

381

0

1

11

Chris Offner

@chrisoffner3d

2 months

@cmuratori I say that as a current CS MSc student but the problem is that knowing somebody has a CS degree gives you very little information. You can meet three people with CS degrees and it could be that person 1 had lots of coding and software application engineering courses but virtually…

2

0

10

Chris Offner

@chrisoffner3d

1 month

@CSProfKGD It's funny – I wrote my BSc thesis in theoretical CS, and undergrad algorithms courses are of course all about discrete maths and graph theory. Now in my master's in Visual Computing and ML I use almost none of that. Maybe physics would've been a better undergrad degree for the…

3

1

10

Chris Offner

@chrisoffner3d

2 months

@AntonObukhov1 @TianfuWang2 Trippy 😁

1

10

Chris Offner

@chrisoffner3d

6 months

Academia: “We propose the first Model that predicts the 3D model of an object from a single input image within just 1 second*.” … *) when using five NVIDIA A100 GPUs costing $50,000 #NeRF

0

10

Chris Offner

@chrisoffner3d

10 days

Synthetic data allows us to get highly realistic RGB, and perfect depth, normals, motion vectors, and any other geometric ground truth data we may want for 3D (or 2D) computer vision. I'd love to see the CV and CG communities combine into "Visual Computing" (my MSc major). 👁️❤️👾

2

0

9

Chris Offner

@chrisoffner3d

1 month

😊

1

0

9

Chris Offner

@chrisoffner3d

5 months

@arankomatsuzaki @giffmana What’s an AI app without fake demos?

0

1

7

Chris Offner

@chrisoffner3d

2 months

The paper vs. the GitHub repository. 🫠

1

0

8

Chris Offner

@chrisoffner3d

5 months

And then the attention maps for all kinds of unrelated pixels also keep obsessing over it. 😱 Stop staring at the corner of my eye, it's impolite!

2

0

8

Chris Offner

@chrisoffner3d

6 days

I'm reading a dissertation where virtually every single sentence is written in the passive voice, and sentences are way too long. Telltale signs that the author is a native German speaker.

2

0

8

Chris Offner

@chrisoffner3d

6 months

@jon_barron @giffmana @ylecun @geoffreyhinton I believe I agree more with Yann than with Geoff but most of all I think it’s just really unfortunate and counterproductive that “AI risk” is now increasingly synonymous with “extinction risk”. There are huge risks I see with AI (and huge benefits) that are not about extinction.

0

6

Chris Offner

@chrisoffner3d

2 months

@LMSacasas I partially agree. The internet does deplete our will power and, via the many ways it shapes our culture, fragments our time and attention so that it becomes way harder to do anything that would require more activation energy, time, and attention, than mindlessly browsing the…

0

6

Chris Offner

@chrisoffner3d

2 months

@iScienceLuvr Maybe optimising for edginess isn’t the path to maximal intelligence? Glad someone ran that experiment. Now we know. ;)

1

0

6

Chris Offner

@chrisoffner3d

1 month

I plotted the percentage of female bachelor and master students at ETH Zurich in 2022 across the 16 departments.

3

1

8

Chris Offner

@chrisoffner3d

2 months

@osanseviero @giffmana Excuse me, it's called Super Kondi Body Attack / Body Combat.

0

6

Chris Offner

@chrisoffner3d

1 month

Per-frame camera extrinsics self-calibration works like a charm! 🫠🫠🫠

0

6

Chris Offner

@chrisoffner3d

22 days

@random_walker 2022: AGI will be a deity that shall conquer the galaxy. 2023: AGI will replace all humans at everything. 2024: AGI is a program that's useful for many things.

1

0

6

Chris Offner

@chrisoffner3d

5 months

@blazer_adam We'll see. @KBingxin , @AntonObukhov1 et al. just showed how to get high-fidelity depth maps out of SD. For relighting we need surface normals and it's still unclear whether SD learns those. suggests that SD does encode "same plane" relationships, so maybe.

What Does Stable Diffusion Know about the 3D Scene?

Recent advances in generative models like Stable Diffusion enable the generation of highly photo-realistic images. Our objective in this paper is to probe the diffusion network to determine to...

arxiv.org

0

5

Chris Offner

@chrisoffner3d

18 days

US politics is so utterly confused. I’m not a fan of TikTok but this is such embarrassing protectionism, lobbied for by Meta. Pass a universal data privacy law that affects US social media companies equally, or get off your high horse.

2

3

6

Chris Offner

@chrisoffner3d

2 months

@jon_barron Yeah, control is the big one here. Waiting (or working) many hours for seconds of final footage is standard procedure even in live action films, not just Pixar animation. People have no idea how long it takes to shoot a scene that ends up being 60 seconds in the final film/show.

1

0

6

Chris Offner

@chrisoffner3d

1 month

I implemented image deformation using moving least squares. Obviously this was the only correct image to test it with.

0

6

Chris Offner

@chrisoffner3d

6 months

Very cool new paper from #ETHZurich : Exponentially Faster Language Modeling

0

3

6

Chris Offner

@chrisoffner3d

5 months

AGI may be near but PDF parsing is still beyond humanity's technological capabilities. Here's how Google's #NotebookLM parses a two-column paper. 🥲 #Bard #Gemini #GoogleGemini #AI

1

6

Chris Offner

@chrisoffner3d

1 month

@m_schuetz If I became supreme dictator of the world, I would mandate a single coordinate system convention and then resign.

2

0

6

Chris Offner

@chrisoffner3d

7 months

Here's a brief excerpt where I remind everyone what Gaussians look like. ;)

1

0

6

Chris Offner

@chrisoffner3d

1 month

@cocktailpeanut @_akhaliq Spread the word! ❤️

Chris Offner

@chrisoffner3d

2 months

Every student with only a MacBook to work on will love you for adding an MPS check to your PyTorch device assignment. 🙃

0

4

31

2

0

5

Chris Offner

@chrisoffner3d

10 days

When vision people do graphics, they call it "image synthesis." When graphics people do vision, they call it "inverse rendering". ;)

0

6

Chris Offner

@chrisoffner3d

6 months

@3DVconf The all-seeing eyes of Hartley & Zisserman.

1

0

6

Chris Offner

@chrisoffner3d

5 months

It's not just my face. Does anybody know the significance of these distinct spots in Stable Diffusion's VAE latents? #StableDiffusion #Diffusion #VAE

2

0

5

Chris Offner

@chrisoffner3d

2 months

One nice thing about studying at ETH is when physicists who also happen to be world-class triathletes casually tell you that “a non-productive day is not the end of the world.” 🫠

1

0

6

Chris Offner

@chrisoffner3d

1 month

I hate everything about this. 😭

Matthew Zeitlin

@MattZeitlin

2 months

in many sectors within like six months fully human writing will be a novelty

53

584

8K

1

0

6

Chris Offner

@chrisoffner3d

2 months

Marigold has such a pretty colourmap for its depth maps. I much prefer the red-to-blue gradient over the more commonly used white-to-black gradient for depth.

0

6

Chris Offner

@chrisoffner3d

20 days

@StefanPijnacker @j_bikker Would love to see those in motion! :)

0

5

Chris Offner

@chrisoffner3d

28 days

@mayfer @kchonyc @LightningAI I use PyTorch on Apple Silicon. 🥹

0

5

Chris Offner

@chrisoffner3d

6 days

@andrewgwils Industry gives greater rewards for doing things well. Academia gives greater rewards for doing things first.

1

0

5

Chris Offner

@chrisoffner3d

5 months

Does Stable Diffusion's VAE encoder think something is wrong with the corner of my eye? 🥺

1

0

4

Chris Offner

@chrisoffner3d

5 months

@CSProfKGD This was part of my initial exploration for our deep learning course project this semester. I’ll have a project report by the end of the semester. Just thought I’d share this because I found it cool. I’m new to DL (and research projects), so don’t know if it had been done before.

0

1

5

Chris Offner

@chrisoffner3d

9 days

Fantastic TED talk by @profgalloway . The median age in the US is ~38.5 years. I wish someone would give this talk in Germany where the median age is 44.9 years.

How the US Is Destroying Young People’s Future | Scott Galloway | TED

In a scorching talk, marketing professor and podcaster Scott Galloway dissects the data showing that, by many measures, young people in the US are worse off ...

www.youtube.com

1

5

Chris Offner

@chrisoffner3d

5 months

I created this graphic to give myself an easily glanceable overview of the U-Net in the Stable Diffusion architecture. Maybe it can be useful to others as well. #StableDiffusion #diffusion #LDM

0

5

Chris Offner

@chrisoffner3d

5 months

@AntonObukhov1 Wow, those depth maps look amazing. Great work! There goes my course project for Deep Learning. 😅

Chris Offner

@chrisoffner3d

6 months

Playing with self-attention in latent diffusion models. This animation illustrates that the model learns to represent 3d scene properties like depth and object semantics. The red dot shows which pixel's self-attention map we're seeing. #Diffusion #DeepLearning #StableDiffusion

13

26

284

0

5

Chris Offner

@chrisoffner3d

10 months

@docmilanfar @mmbronstein "Müller" is the most common German surname. "Thomas Müller" is to Germans what "John Smith" is to Brits.

1

0

4

Chris Offner

@chrisoffner3d

13 days

"The proton is a quantum mechanical object that exists as a haze of probabilities until an experiment forces it to take a concrete form. And its forms differ drastically depending on how researchers set up their experiment." Sophon lock confirmed.

Inside the Proton, the ‘Most Complicated Thing’ Imaginable

The positively charged particle at the heart of the atom is an object of unspeakable complexity, one that changes its appearance depending on how it is probed. We’ve attempted to connect the proton’s...

www.quantamagazine.org

0

3

5

Chris Offner

@chrisoffner3d

2 months

The nighttime view from Polyterasse at ETH Zurich.

0

4

Chris Offner

@chrisoffner3d

2 months

Nice doggo statue in Lucerne, scanned with @LumaLabsAI .

0

5

Chris Offner

@chrisoffner3d

7 months

@nathanbenaich @RikeFranke @ylecun The saying "He has forgotten more about X than I'll ever know." means that a person is such an expert that even the volume of X that he has forgotten exceed the speaker's current knowledge of X. It's not diminishing LeCun's expertise, it's complimenting it.

0

5

Chris Offner

@chrisoffner3d

1 month

@yongyuanxi @ZGojcic The more time I spend among academic researchers, the more I understand why the German word for “curiosity” is “Neugier”, as in “greed for novelty” (neu = new, Gier = greed). I also notice it in myself. I like that you point out the tension between self-serving and…

0

4

Chris Offner

@chrisoffner3d

2 months

If Twitter could just allow me to auto-block any user (bot) profile that has a link to a camgirl site in its bio, that would go a long way. I just purged my follower list of ~40 obvious bot accounts and I'm sure I've missed many more.

2

1

5

Chris Offner

@chrisoffner3d

7 months

Rotating the covariance matrix of a 3D Gaussian and sampling points from it each frame.

0

5

Chris Offner

@chrisoffner3d

1 month

@JustinMSolomon @sp_monte_carlo I believe it's just a remnant of the standard DL introduction where you go from a Perceptron to an MLP, and these computational graph diagrams then justify the term neural "network." I also really don't think that narrative is worth it though.

0

5

Chris Offner

@chrisoffner3d

6 months

Cool example of neural rendering using a model trained specifically for this game. Inputs are ball, paddle, and light positions (7 floats), output is the per-object texture (256x256). Training data created with Blender cycles.

From the learnmachinelearning community on Reddit: A simple Neural Rendering demo running Ray...

Explore this post and more from the learnmachinelearning community

www.reddit.com

Immortal Ray

@immortalray98

6 months

Ray Traced Pong starting to looking pretty good on integrated graphics (neural rendering using @unity #Sentis ) #MachineLearning #unity #madewithunity

0

1

12

1

0

5

Chris Offner

@chrisoffner3d

21 days

@YiMaTweets Did it ask you for your clothes, your boots, and your motorcycle?

2

0

5

Chris Offner

@chrisoffner3d

6 months

I wonder if universities could create a custom GPT for each course to provide students with a tutor. Supply the GPT with all course materials, related readings, problem sets, and instructions, and let students use it as a 24/7 available TA. #education #chatGPT #GPTs #openAI

3

0

5

Chris Offner

@chrisoffner3d

26 days

Hell yeah! 👾

1

0

5

Chris Offner

@chrisoffner3d

6 months

@YiMaTweets @young_opsimath I agree in principle but in practice unfortunately not all lines of inquiry lend themselves to rigorous experiments. And logical deduction reaches its limits pretty quickly as the complexity and stochasticity of the system under observation increases.

1

0

5

Chris Offner

@chrisoffner3d

17 days

@Parskatt

0

1

5

Chris Offner

@chrisoffner3d

27 days

@CSProfKGD Seems like a common experience.

Justin Solomon

@JustinMSolomon

1 month

Thank you to the authors for a detailed response that addresses all my concerns and provides the experiments I requested. I have no further questions or comments and choose to leave my score the same.

6

7

115

2

0

4

Chris Offner

@chrisoffner3d

7 months

More interesting: playing with the parameters of a 2D Gaussian.

2

0

4

Chris Offner

@chrisoffner3d

5 months

Me studying ML maths.

0

4

Chris Offner

@chrisoffner3d

3 months

@Peter_shirley The "Computer Graphics" course at ETH Zurich is really more of a "Physically Based Rendering" course, and uses the pbrt book.

1

0

3

Chris Offner

@chrisoffner3d

2 months

@docmilanfar Sir, could you please spare some lecture titles? 😅

0

4

Chris Offner

@chrisoffner3d

1 month

On 19.4.2013, when ETH Zurich's president at the time, Prof. Dr. Ralph Eichler, was asked about how to increase the ratio of female students, this was his answer: "I think after a few years I better understand the problem. Women have a much greater need for a sense of purpose.…

3

0

4

Chris Offner

@chrisoffner3d

8 days

What People Think I Do / What I Really Do with @clairebraeuer 😅❤️

1

4

Chris Offner

@chrisoffner3d

2 months

@JonasAndrulis This is one of the greatest issues I see with rapid progress – it leaves us in constant analysis paralysis about what long-term projects are even worth starting and leads to short-term thinking. If the rules change all the time, how can we hope to make sensible long-term plans?

2

0

3

Chris Offner

@chrisoffner3d

28 days

CREStereo performs pretty nicely but has some trouble with fine details in the background.

1

0

4

Chris Offner

@chrisoffner3d

10 days

@AlbyHojel @JeromeRevaud @Parskatt My background shapes my biases here but I think some ambitious (3D) vision group should hire some CG/VFX folks who understand the needs of CV research, and have them create the greatest synthetic 3D vision dataset ever seen. 🙃 Happy to talk in DMs if anyone is interested.

0

4

Chris Offner

@chrisoffner3d

12 days

It doesn't help that the main business pitch for AI/ML is that it can/will automate away much of what people currently earn their living with. Maybe once enough people flee into ML, the AI crowd will learn what it means to have their market destroyed by cheap and abundant labour.

Amir-massoud Farahmand

@SoloGen

13 days

Most people nowadays are interested in AI/ML because of the job prospective, or because it may lead to the growth of capital. These are people who either hope to learn about ML so that ... (2/10)

1

0

5

0

4

Chris Offner

@chrisoffner3d

4 months

Cross-attention maps in Stable Diffusion's U-Net for time steps t = 1, ..., 1000.

0

3

Chris Offner

@chrisoffner3d

1 month

@ducha_aiki @Parskatt @janusch_patas @marcorsingher @SattlerTorsten Does anyone truly understand CMake? 😅

1

0

4

Chris Offner

@chrisoffner3d

2 months

@ftm_guney I apologise for perpetuating unrealistic beauty standards for stereo matchers. 😅 But it seems useful to first check “How would (self-)calibration method X perform with perfect matches?” and only once something works there, move on to the realistic case of imperfect matches.

0

3

Chris Offner

@chrisoffner3d

15 days

I'm always surprised when I meet Germans who in 2024 still hold a dismissive view of Chinese manufacturing, unaware that Germany's path to prosperity looked very similar. "Made in China" may soon be as respected as "Made in Germany."

1

0

4

Chris Offner

@chrisoffner3d

1 month

@BarmanPalak @stephmui Then why are they dating techies? ;)

0

4

Chris Offner

@chrisoffner3d

2 months

@Michaellaskey7 I don’t disagree with the tweet on its own but the video seems like an odd illustration of that claim. Grass lawns and lawnmowers have little to do with “the environment” or sustainability.

1

0

3

Chris Offner

@chrisoffner3d

5 months

This is brought to you by our Reinforcement Learning homework in which we were asked to implement the SAC algorithm. This was maybe not the optimal choice for literally the first Reinforcement Learning assignment given to students. 🫠 #ReinforcementLearning #RL

1

0

4

Chris Offner

@chrisoffner3d

23 days

@jenwiderberg @thealexbanks Making sure you can never talk to a human being for customer service again.

0

4

Chris Offner

@chrisoffner3d

5 months

As a German, I thank ETH for providing an emerging country like Germany with such generous development aid in its digital transformation efforts. 😅

ETH Zurich

@ETH_en

5 months

ETH and the Dieter Schwarz Foundation have signed a far-reaching letter of intent to establish a new teaching and research centre for responsible digital transformation in Heilbronn. Over the next 30 years, around 20 professorships shall be established. 🇩🇪

8

83

0

4