moetsi

Reality Modeling as a Service™. We build a 1:1 representation of physical reality, in real-time.

[Matthias Niessner @MattNiessner]

Check out TriPlaneNet!

From a single image, we predict EG3D latents & offsets, thus obtaining high-fidelity 3D models. Works in real time!

Great work by our stellar MA student @anantarbb advised by @ASevastopolsky #WACV2024

https://t.co/0PNl26iaDW
https://t.co/45h21TF27N

[Angela Dai @angelaqdai]

We've released the ScanNet++ data!
Check it out: https://t.co/SKCGM23hA0
280 high-fidelity 3D scenes w/ 1mm geometry, DSLR+iPhone images, semantics

We're currently beta-testing, please bear with us - approval may initially take up to 2 weeks

Test scenes and benchmark to come!

[Alberto @taiuti]

Mapbox has a brand new 3D map style, still in beta!
But it didn’t have a hackable live demo, so at Skydio we created one 😎 , try it out here: https://t.co/EG2f50hLjK

The updates include:
- Hundreds of new 3D models for famous landmarks
- New symbolic design that works hand in hand with the style
- Day/night light shift built in!

How does this help? Through more accurate 3D maps:
- Data displayed on the map is more useful as it’s easier to understand where you are (our brains think in 3D)
- More immersive 3D experiences are unlocked, not just navigation

Looking forward to seeing where @Mapbox takes this next

https://t.co/ducGuBnp30

[Bilawal Sidhu @bilawalsidhu]

NVIDIA's Omniverse was big news in Jensen Huang's SIGGRAPH 2023 keynote today.

Here's everything you need to know:

Omniverse is a 3D collaboration platform that enables the creation of massive virtual worlds with photorealistic graphics and (now) AI-generated content.

NVIDIA wants Omniverse to be the place where AI can learn in a virtual world... but also the place where AI can help you create virtual worlds.

Literally connecting bits and atoms -- NVIDIA wants to help the 50 trillion dollar heavy industry that is racing to "digitalize" their operations.

Key new capabilities announced:
• Integration of generative AI like Adobe Firefly to enhance creation workflows (wow!)
• Expanded ecosystem connections through OpenUSD (Adobe, Wonder Dynamics, Luma AI, etc)
• New developer tools and templates for building apps and experiences
• Semantic search capability with DeepSearch to find 3D assets easily
• Optimizations for photorealistic real-time rendering and path tracing with AI-accelerated denoising powered by new RTX GPUs
• XR capabilities native to the platform (so you can deploy on AR/VR headsets)
• Upgrades to core apps like Omniverse Audio2Face and USD Composer
• Graphics Delivery Network (GDN) to performantly serve your 3D experience around the world
• Support for new workflows across industrial use cases like digital twins

With this release, there's a focus on breaking data silos and connecting complex CAD and 3D pipelines enabled by the OpenUSD format. The applications are limitless - product design, digital twins, virtual production, training simulations, and more!

I also had the chance to sit down with @RevLebaredian, NVIDIA's VP of Omniverse and Simulation Technologies. Follow me @bilawalsidhu and stay tuned in the coming days for that deep dive interview!

[moetsi @moetsi6]

@MakersMark 10000000% would buy it

[Jeff Dean @JeffDean]

The Open Buildings V3 dataset of 1.8B building detections w/higher precision & recall is now available from@GoogleResearch, adding coverage for Latin America and the Caribbean (in addition to coverage of Africa and South and Southeast Asia from V1/V2).

https://t.co/CyPStYdH51

[moetsi @moetsi6]

new video from @LearnOpenCV 👇 about "Understanding Intersection Over Union (IoU)" in object detection & segmentation

https://t.co/SUABAHTo0h

#computervision #opencv #spatialcomputing

[Zhenjun Zhao @zhenjun_zhao]

AnyLoc: Towards Universal Visual Place Recognition

Nikhil Keetha, Avneesh Mishra, Jay Karhade, Krishna Murthy Jatavallabhula, Sebastian Scherer, Madhava Krishna, Sourav Garg

tl;dr: foundation model features+unsupervised feature aggregation

https://t.co/nOIn6CsetE

[International Conference on 3D Vision (3DV) @3DVconf]

Best Paper Honorable Mention at #3DV2017

Fast Incremental Bundle Adjustment with Covariance Recovery

Viorela Ila, Lukas Polok, Marek Solony, Klemen Istenic

https://t.co/2rafZk3H4p

#TBThursday #3DV2024

[moetsi @moetsi6]

congrats to @MattNiessner & team!

#ICCV2023

[Matthias Niessner @MattNiessner]

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 :)

https://t.co/rEHOJusF4g

[Yinyu Nie @yinyu_nie]

Being aware of many students taking a long time in rendering and visualizing the 3D-Front dataset. Here I cleaned and organized my version, and now it supports Multi-GPU batch rendering. Hope it helps.

Github: https://t.co/6iU018BUxb

[Andrew Davison @AjdDavison]

This looks very interesting.
@HideMatsu82

[moetsi @moetsi6]

congrats to @GKopanas 👏👏👏

#SIGGRAPH2023 #computervision #spatialcomputing

[Andrea Tagliasacchi @taiyasaki]

Congrats for the #SIGGRAPH2023 best paper award @GKopanas 🎉🎉

[moetsi @moetsi6]

great stuff @CixLiv

#computervision #AR #VisionPro

[CIX 🦾 @cixliv]

Today I announce my newest thing! @VENDOR_GG!

"Sell items in seconds using AI"

Here showing a concept with VisionOS (Apple Vision Pro).

Launching on iOS + Discord Bot later this month for SF!

Super excited to be founding with @mentarus a @ycombinator founder and good friend.

[moetsi @moetsi6]

👏👏

#spatialcomputing #sensors #lidar

[Zhenjun Zhao @zhenjun_zhao]

UnLoc: A Universal Localization Method for Autonomous Vehicles using LiDAR, Radar and/or Camera Input

Muhammad Ibrahim, Naveed Akhtar, Saeed Anwar, Ajmal Mian

tl;dr: multi modality localization

https://t.co/NzrYDJ56Ic

[Satya Mallick @LearnOpenCV]

🎥New Video:
https://t.co/MfidGJzK7w

Dive into the specifics of SAM in our latest video - a game-changing image segmentation model with unmatched precision. #AI #ImageSegmentation #AIInnovation #computervision

[moetsi @moetsi6]

incredible name 😂 The Drunkard's Odometry!
#computervision #dataset @zhenjun_zhao

[Zhenjun Zhao @zhenjun_zhao]

The Drunkard's Odometry: Estimating Camera Motion in Deforming Scenes

David Recasens, @Martin_R_Oswald , @mapo1 , @jcivera

tl;dr: dataset in deformable environments; deformable odometry

https://t.co/lNISqaM4mr
https://t.co/1iIwStQINU

[dmvrg @dmvrg]

video streaming in #AR with #AI generated depth map, spatial Netflix

[moetsi @moetsi6]

yay the OAK-D-SR is here!

that measurement accuracy 🤤

cc @luxonis

#computervision #spatialcomputing #sensors

[Luxonis | Robotic Vision @luxonis]

📣Meet the OAK-D-SR camera from Luxonis, an innovative solution with excellent close-up measurement accuracy📏, competitive pricing💰, and a compact design📷. Perfect for robotics and detailed measurements. Explore our blog for more.

https://t.co/YdF4xfiPE2

[moetsi @moetsi6]

incredible stuff @MattNiessner @OHilliges

RGB and not RGB-D!

[Matthias Niessner @MattNiessner]

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.

https://t.co/PzQx3qnNIj