Opens profile photo
Follow
Click to Follow VGolyanik
Vlad Golyanik
@VGolyanik
Research group leader at MPI for Informatics working on methods for 3D reconstruction of deformable scenes and quantum computer vision.
Saarbrücken, Germanypeople.mpi-inf.mpg.de/~golyanik/Joined August 2021

Vlad Golyanik’s Tweets

Can rf-SQUID qubits be used to generate random numbers? It is simple in theory (=initialise and measure), but what do we observe in practice? See our study published in IEEE Access on the generation of truly random numbers on D-Wave quantum annealers: 4dqv.mpi-inf.mpg.de/QRNG/
Embedded video
0:16
75 views
3
Concurrent studies are not uncommon in computer vision 🎭 What about other fields of science? Well... The Higgs boson was predicted in three independently published papers more than half a century ago, and there are much older examples in history!
Image
Image
Image
1
😱 AlphaTensor. 😱 This. Is. Huge. Reinventing matrix multiplication using Deep RL! Faster than Strassen's algorithm that has stood the test of time for 50+ years! 10-20% efficiency increase! Imagine what that translates to considering how fundamental matrix multiplication.. 1/
Quote Tweet
Today in @Nature: #AlphaTensor, an AI system for discovering novel, efficient, and exact algorithms for matrix multiplication - a building block of modern computations. AlphaTensor finds faster algorithms for many matrix sizes: dpmd.ai/dm-alpha-tensor & dpmd.ai/nature-alpha-t 1/
Show this thread
Embedded video
0:07
737K views
6
649
Show this thread
Congratulations to #Svante Pääbo, director @ the Max Planck Institute for Evolutionary Anthropology on winning this year's #NobelPrize in Physiology or Medicine! 🙃😁🙃😃🤗 😃🤗🍾🍾🤗🍾🤗🙃
Quote Tweet
BREAKING NEWS: The 2022 #NobelPrize in Physiology or Medicine has been awarded to Svante Pääbo “for his discoveries concerning the genomes of extinct hominins and human evolution.”
Image
9
1,541
Everyone knows that all circles are similar. But did you know that all parabolas are similar? The ratio of the red arc and the blue focal segment is √2 + log(1+√2) = 2.29558... for every parabola. This is the universal parabolic constant, the “π of parabolas”.
Image
31
2,228
UnrealEgo is both a new dataset AND a new method for egocentric 3D human pose estimation. The paper will be presented at #ECCV2022 (main conference) and workshop ego4d-data.org/workshops/eccv held in conjunction with #ECCV2022 in Tel Aviv. Project page: 4dqv.mpi-inf.mpg.de/UnrealEgo/
Quote Tweet
We also introduce a new method for 3D human pose estimation achieving high accuracy on UnrealEgo. Is UnrealEgo also the first open-sourced approach for egocentric 3D human body reconstruction?.. Check the source code here: github.com/hiroyasuakada/.
Show this thread
Embedded video
0:43
203 views
17
Characteristics of UnrealEgo: • Naturalistic • High character diversity and motion complexity • The characters and backgrounds are modelled in 3D • Provides stereo views from an eyeglass frame • The Unreal Engine files allow modifications.
1
1
Show this thread
The Navier-Stokes Equations 💦 In 1845, Sir George Stokes had derived the equation of motion of a viscous flow by adding Newtonian viscous terms, thereby the NSEs had been brought to their final form which has been used to generate numerical solutions for fluid flow ever since.
Image
10
1,251
Show this thread
The core motivation for the joint work with , M. Elgharib, , and C. Theobalt was semantic control over the learnt appearance thanks to the intrinsic decomposition. It enables even surreal relighting! Full video: youtu.be/V8h77YMcxTg
Quote Tweet
A new #ECCV2022 paper, NeRF for Outdoor Scene Relighting, is out! Project Page: 4dqv.mpi-inf.mpg.de/NeRF-OSR/ Paper: arxiv.org/pdf/2112.05140 Source Code: github.com/r00tman/NeRF-O New benchmark dataset for outdoor scene relighting with eight sites (33GB): nextcloud.mpi-klsb.mpg.de/index.php/s/mG
Embedded video
0:17
10.5K views
9
The source code of phi-SfT is now available: github.com/navamikairanda Join our virtual presentation at #CVPR2022 that will take place tomorrow at 10 am and 10 pm CT or 5 pm and 5 am (next day) GMT+2.
Quote Tweet
A new shape-from-template approach for dense monocular non-rigid 3D reconstruction of challenging cloth deformations; to be presented at #CVPR2022. Key components: a differentiable renderer and a differentiable physics simulator. 4dqv.mpi-inf.mpg.de/phi-SfT/ arxiv.org/abs/2203.11938
Embedded video
0:34
444 views
6
Finally! Check the extended version of our state of the art report (STAR) talk on Advances in Neural Rendering presented at #EUROGRAPHICS2022! Many thanks to all speakers and co-authors! [EUROGRAPHICS 2022, STAR] Advances in Neural Rendering youtu.be/ul9hFFtWYv8 via
4