We will organise a workshop on Quantum Computer Vision and Machine Learning at #CVPR2023. Will the future be quantum? Stay entangled!
Pinned Tweet
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.
Vlad Golyanik’s Tweets
Oldies but goldies: Immo Kerner, Ein Gesamtschrittverfahren zur Berechnung der Nullstellen von Polynomen, 1966. The Durand–Kerner method is surprisingly simple way of computing all the roots of a polynomial. en.wikipedia.org/wiki/Durand%E2
0:05
25.3K views
6
81
503
Our QuAnt paper on QUBO learning has been accepted at #ICLR 2023 as a spotlight! Instead of deriving QUBO formulations for modern quantum annealers analytically, we learn the qubit biases and couplings (QUBO instantiation) for target problems from data.
arxiv.org/pdf/2210.08114
0:19
433 views
1
3
14
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/
0:16
75 views
3
BREAKING NEWS:
The Royal Swedish Academy of Sciences has decided to award the 2022 #NobelPrize in Physics to Alain Aspect, John F. Clauser and Anton Zeilinger.
420
15K
33.8K
Show this thread
Serving as a reviewer for #CVPR2023, check out this #CVPR2020 tutorial on “How to write a good review?”
youtu.be/W1zPtTt43LI
2
19
104
Show this thread
I’ve just released mesh-to-sdf for Unity!
A light and fast real-time SDF generator, primarily for animated characters.
The dynamic SDF output can be used for all sorts of VFX 💥 It also enables hair-to-character collisions in the new hair package.
github.com/Unity-Technolo
0:23
42.3K views
26
372
1,911
Show this thread
In 1972, John Conway posed the following question with an associated $50 cash prize: in Game of Life, is there a stable configuration whose only father is itself?
Törmä and Salo found a solution in January 2022.
28
374
3,154
Neural Volume Rendering for Dynamic Scenes
NeRF has shown incredible view synthesis results, but it requires multi-view captures for STATIC scenes.
How can we achieve view synthesis for DYNAMIC scenes from a single video? Here is what I learned from several recent efforts.
0:03
10.2K views
5
78
328
Show this thread
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!
1
This was an outstanding, well-organised and fun in Tel Aviv! Many thanks to all (the eccv team and the attendees) for making it happen! 👏👏👏
27
1
15
Our paper got accepted by #BMVC2022!
HiFECap is a novel neural human performance capture approach simultaneously capturing human poses, clothing deformation, facial expressions, and hand gestures just from a single RGB video.
people.mpi-inf.mpg.de/~mhaberma/proj
0:18
1.1K views
2
7
57
😱 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
0:07
737K views
6
156
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.”
9
255
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”.
31
399
2,228
2
9
Japanese Stable Diffusion
using approximately 100 million images with Japanese captions, including the Japanese subset of LAION-5B to train the model
model card: huggingface.co/rinna/japanese
github: github.com/rinnakk/japane
colab: colab.research.google.com/drive/1i2el5qh
3
80
310
Show this thread
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
0:43
203 views
4
17
Adiabatic quantum computers such as , specialize in quadratic unconstrained binary optimization (QUBO). Yet, real applications come with constraints. Our #ECCV2022 work [arxiv.org/abs/2203.12633] devises the first hybrid solver for linearly constrained QBO problems. 1/4
1
1
13
Show this thread
Our new MoCapDeform approach for scene-aware monocular 3D human motion capture achieves superior accuracy on different metrics by modelling 3D scene deformations that related works ignore.
#3DV2022 (Oral)
Project page: 4dqv.mpi-inf.mpg.de/MoCapDeform/
Paper: arxiv.org/pdf/2208.08439
0:47
488 views
2
5
24
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/.
0:43
203 views
1
1
Show this thread
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
Thanks to all my co-authors from and MPI for Informatics!
Project page: 4dqv.mpi-inf.mpg.de/UnrealEgo/
Paper: arxiv.org/pdf/2208.01633
The dataset (~2.2TB) and the data generation tools (~20MB) are available.
2
1
Show this thread
What was missing in egocentric 3D human motion capture so far? A large-scale dataset with high motion complexity and character diversity, we believe.
Today we are excited to announce 's UnrealEgo dataset accepted at #ECCV2022.
0:34
1.2K views
1
6
34
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.
10
290
1,251
Show this thread
Time to reinvent the wheel.
Here’s the Cartwheel Galaxy in a whole new light — as a composite image from 2 instruments on the Webb telescope. Webb uniquely offers not just a snapshot of the galaxy’s current state, but also a peek into its past & future: go.nasa.gov/3SfEQgR
read image description
ALT
404
9,176
39.8K
Show this thread
Meet Soshi Shimada's HULC 🧌 at #ECCV2022, the new SOTA in scene-aware🛋️🪑🚶 global 3D human motion capture. The dataset (GTA-IM🏎️) with dense contact annotations is available on the project page: vcai.mpi-inf.mpg.de/projects/HULC/.
PS: 🥷Yes, we are following the naming convention.
0:38
1.3K views
10
49
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
0:17
10.5K views
9
QuMoSeg (#ECCV2022, 4dqv.mpi-inf.mpg.de/QuMoSeg/) is a new quantum-classical approach for motion segmentation demonstrating the advantages of qubits and quantum hardware in computer vision! Many thanks to my co-authors , , and !
0:22
194 views
3
8
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
0:17
10.5K views
3
77
294
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
0:34
444 views
2
6
Physical Inertial Poser (arxiv.org/pdf/2203.08528) eventually made it to the list of 33 best paper finalists at #CVPR2022.
PIP combines a neural kinematics estimator and a physics-aware motion optimiser for 3D human body tracking with six IMUs.
1
8
Physics-based constraints significantly improve the accuracy of monocular 3D human motion capture. And what about IMU-based 3D mocap? Check Xinyu Yi et al.'s #CVPR2022 paper Physical Inertial Poser (PIP) for the answer:
xinyu-yi.github.io/PIP/
arxiv.org/pdf/2203.08528
0:39
270 views
1
2
10
The source code of Neural PhysCap by Shimada et al.("Neural Monocular 3D Human Motion Capture with Physical Awareness") presented at #SIGGRAPH2021 is now available at #github!
Project page: vcai.mpi-inf.mpg.de/projects/PhysA
Source code: github.com/soshishimada/N
0:31
135 views
1
8
This is a joint work with , , Aliaksandr Siarohin, Christian Theobalt, and .
youtu.be/FuIulCRh_hY via
1
3
Show this thread
#CVPR2022: Our Playable Environments is a generative, compositional and controllable (by learnt actions) neural volumetric 3D model that allows rendering videos with the camera, agent and style control from a single monocular image.
arxiv.org/pdf/2203.01914
willi-menapace.github.io/playable-envir
0:30
498 views
3
3
19
Show this thread
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
2
4