kaplanyan

Excited to share 2025 update from our team, including our presence at #HPG and #SIGGRAPH: community.intel.com/t5/Blogs/T
We focus on visual efficiency and democratizing high-end visual experiences

#DirectX Cooperative Vectors are live: devblogs.microsoft.com/directx This unlocks our systolic hardware. We provide a sample for Intel discrete and integrated GPUs. This is an exciting push across hardware vendors and #Microsoft to unleash the new era of ML-assisted visuals

Attending #GDC? Neural block texture compression is ~5x smaller than BC6, can run with practical performance on Intel GPUs and will be available for game developers.
Join us Monday morning for a joint talk with Microsoft on accelerated neural graphics

#SIGGRAPH it’s this time of the year. We will show a demo of Nd Gaussian splats and have a talk at Natasha’s course after the HPG paper. Looking forward to meeting the community in person again. Lots of exciting content

What it means in practice? It enables an efficient representation of dynamic scenes: animations, dynamic objects, moving lights etc.

Check out our SIGGRAPH’24 work on why Nd Gaussians are a good idea. Gaussians are easier than NeRF to optimize. They have localized mass and gradients for isolated placement of density. Important for higher dimensions, they also start sparse, unlike MLPs. community.intel.com/t5/Blogs/T

kaplanyan boosted:

Registration for REAC 2024 is now open! Check our page to see the amazing speaker lineup we have this year! enginearchitecture.org/2024.ht - and please share!

@BartWronski another factor is public schools, but might be a bigger thing on the west coast

@aras all the credit goes to the folks :) Well deserved. Hard work, nice presentation, pleasure to read. You’ve probably seen them in action before :)

As a result of HPG 2023, Thomas Deliot and Laurent Belcour received the 1st place award for their real-time glints work.

And Jonathan Dupuy, and Anis Benioub received the 2nd place award for their work GGX sampling. Congratulations on the amazing and hard work!

Intel Graphics Research presents 8 papers at SIGGRAPH/EGSR/HPG and will give a talk at Advances in Real-time Rendering course. Low-energy path tracing, generative AI, neural graphics and more! Amazing work of our researchers and collaborators.

intel.com/content/www/us/en/de

GGX team simplifies the concept of diffusion models to make it acceccible to more people around the world. As a bonus, it makes a connection to well-known graphics concepts. Check our this new SIGGRAPH’23 work as a good way to start your journey in diffusion models intel.com/content/www/us/en/de

Our #Siggraph23 paper demonstrating the example of deep integration of machine learning into modern rendering pipeline. Here is a more detailed readup: intel.com/content/www/us/en/de

Not sure how many MSc/PhD students are here, just in case, Computer Graphics@KIT (my alma mater) is looking for PhD students on realistic material models in real-time path tracing, as well as PhD candidates/PostDocs for rendering/visualization research cg.ivd.kit.edu/2033.php can highly recommend the lab, lots of fun!

Thanks for inviting Marc and Patric! It’s an honor. I really enjoyed our fireside chat on furute opportunities at the intersection of graphics, GPUs, AI, and immersive visuals. It’s also my first podcast, and I really enjoyed the format
cesium.com/open-metaverse-podc

Looking for a 2023 internship? Intel Graphics Research is hiring in graphics and content creation/generation to advance Intel GPUs. Example topics are neural rendering, distributed rendering, path tracing, differentiable rendering, perceptual graphics. E-mail our scientist you know or want to work with from here intel.com/content/www/us/en/de Or e-mail me. Locations include Europe and US

@castano a few other things to consider. The signal you’re compressing can be not linear in error, for example, normal maps. Worth considering a different error function. And then there is also cross-block error optimization to tradeoff total block error for blockiness

Client Info

Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst