Marco Salvi
@marcosalvi.bsky.social
Computationally responsible deep learning and real-time graphics research @ NVIDIA. Personal opinions only.
Couldn't be prouder of my intern @bingxxu.bsky.social
for leading this work! 🚀
It took immense persistence to turn the initial hunch, that an attention mechanism could model long-range interactions in light transport, into a robust, working system. Huge congrats to the entire team! 🎉
for leading this work! 🚀
It took immense persistence to turn the initial hunch, that an attention mechanism could model long-range interactions in light transport, into a robust, working system. Huge congrats to the entire team! 🎉
🎄 Introducing our paper A Generalizable Light Transport 3D Embedding for Global Illumination lnkd.in/gQUMSAyV .
🙈 Just as Transformers learn long-range relationships between words or pixels, our new paper shows they can also learn how light interacts and bounces around a 3D scene.
🙈 Just as Transformers learn long-range relationships between words or pixels, our new paper shows they can also learn how light interacts and bounces around a 3D scene.
October 23, 2025 at 3:40 AM
Couldn't be prouder of my intern @bingxxu.bsky.social
for leading this work! 🚀
It took immense persistence to turn the initial hunch, that an attention mechanism could model long-range interactions in light transport, into a robust, working system. Huge congrats to the entire team! 🎉
for leading this work! 🚀
It took immense persistence to turn the initial hunch, that an attention mechanism could model long-range interactions in light transport, into a robust, working system. Huge congrats to the entire team! 🎉
Reposted by Marco Salvi
Full replay of the NVIDIA GDC 2025 opening presentation by John Spitzer. Great watch for anyone interested the future of game tech and RTX!👍 www.youtube.com/watch?v=c7PA...
GDC 2025 | Advances in RTX - Full Session Replay
YouTube video by NVIDIA Game Developer
www.youtube.com
April 9, 2025 at 3:17 PM
Full replay of the NVIDIA GDC 2025 opening presentation by John Spitzer. Great watch for anyone interested the future of game tech and RTX!👍 www.youtube.com/watch?v=c7PA...
Reposted by Marco Salvi
To the world:
We are fighting back. Our movement has been silenced by the media here—but we are not backing down. This is what our streets looked like across multiple cities. Tomorrow, there will be more of us! Raise a glass to freedom.
—With love,
Your American allies.
We are fighting back. Our movement has been silenced by the media here—but we are not backing down. This is what our streets looked like across multiple cities. Tomorrow, there will be more of us! Raise a glass to freedom.
—With love,
Your American allies.
April 5, 2025 at 10:03 PM
To the world:
We are fighting back. Our movement has been silenced by the media here—but we are not backing down. This is what our streets looked like across multiple cities. Tomorrow, there will be more of us! Raise a glass to freedom.
—With love,
Your American allies.
We are fighting back. Our movement has been silenced by the media here—but we are not backing down. This is what our streets looked like across multiple cities. Tomorrow, there will be more of us! Raise a glass to freedom.
—With love,
Your American allies.
Reposted by Marco Salvi
Adult in the (cabinet) room:
April 4, 2025 at 2:28 PM
Adult in the (cabinet) room:
I’ve never seen so many fools (including billionaire sycophants..) telling us we don’t get it and that it is all for our own good.
April 4, 2025 at 9:14 PM
I’ve never seen so many fools (including billionaire sycophants..) telling us we don’t get it and that it is all for our own good.
Reposted by Marco Salvi
Sites like google scholar could plausibly improve collaboration by implementing randomized ordering for equal co-authors. Would be pretty cool.
April 2, 2025 at 12:18 AM
Sites like google scholar could plausibly improve collaboration by implementing randomized ordering for equal co-authors. Would be pretty cool.
Reposted by Marco Salvi
zeamoxwang.github.io/HotSpot-CVPR...
My student Zimo/Cheng's recent work tackles this problem! A lot of recent neural "SDF" optimizers use losses that, even when perfectly minimized, still don't result in actual signed distance fields. Our loss guarantees convergence to distance when minimized.
My student Zimo/Cheng's recent work tackles this problem! A lot of recent neural "SDF" optimizers use losses that, even when perfectly minimized, still don't result in actual signed distance fields. Our loss guarantees convergence to distance when minimized.
March 29, 2025 at 9:08 PM
zeamoxwang.github.io/HotSpot-CVPR...
My student Zimo/Cheng's recent work tackles this problem! A lot of recent neural "SDF" optimizers use losses that, even when perfectly minimized, still don't result in actual signed distance fields. Our loss guarantees convergence to distance when minimized.
My student Zimo/Cheng's recent work tackles this problem! A lot of recent neural "SDF" optimizers use losses that, even when perfectly minimized, still don't result in actual signed distance fields. Our loss guarantees convergence to distance when minimized.
Reposted by Marco Salvi
White House says no one will be fired over the Signal scandal because no one involved was trying to cure cancer.
March 24, 2025 at 9:54 PM
White House says no one will be fired over the Signal scandal because no one involved was trying to cure cancer.
Reposted by Marco Salvi
DLSS 3 vs DLSS 4 with GeForce RTX in Assassin's Creed Shadows using DLSS Overrides. The image quality improvement is massive! 🤩
March 23, 2025 at 10:35 PM
DLSS 3 vs DLSS 4 with GeForce RTX in Assassin's Creed Shadows using DLSS Overrides. The image quality improvement is massive! 🤩
Reposted by Marco Salvi
Some highlights from a really compelling overview talk by @lukasheinrich.com this morning at #ML4PS at @neuripsconf.bsky.social -- reflecting on the future of ML for physics, including the importance of multi-modality and when/how to impose our scientific intuitions onto our models.
December 16, 2024 at 12:16 AM
Some highlights from a really compelling overview talk by @lukasheinrich.com this morning at #ML4PS at @neuripsconf.bsky.social -- reflecting on the future of ML for physics, including the importance of multi-modality and when/how to impose our scientific intuitions onto our models.
Reposted by Marco Salvi
Here's a recording of my Tiny Glade presentation from the Graphics Programming Conference 2024:
www.youtube.com/watch?v=jusW...
It's an any% speedrun of our GPU-driven rendering, shadows, global illumination, water, and DoF, involving a few weird tricks.
www.youtube.com/watch?v=jusW...
It's an any% speedrun of our GPU-driven rendering, shadows, global illumination, water, and DoF, involving a few weird tricks.
Rendering Tiny Glades With Entirely Too Much Ray Marching
YouTube video by Graphics Programming Conference
www.youtube.com
December 5, 2024 at 2:19 AM
Here's a recording of my Tiny Glade presentation from the Graphics Programming Conference 2024:
www.youtube.com/watch?v=jusW...
It's an any% speedrun of our GPU-driven rendering, shadows, global illumination, water, and DoF, involving a few weird tricks.
www.youtube.com/watch?v=jusW...
It's an any% speedrun of our GPU-driven rendering, shadows, global illumination, water, and DoF, involving a few weird tricks.
Reposted by Marco Salvi
we've been busy working on 3d world generation at world labs the past few months
today we're sharing an early preview of our technology
check out some interactive results on our website -- worldlabs.ai/blog
today we're sharing an early preview of our technology
check out some interactive results on our website -- worldlabs.ai/blog
December 2, 2024 at 11:06 PM
we've been busy working on 3d world generation at world labs the past few months
today we're sharing an early preview of our technology
check out some interactive results on our website -- worldlabs.ai/blog
today we're sharing an early preview of our technology
check out some interactive results on our website -- worldlabs.ai/blog
Reposted by Marco Salvi
Differentiable physically-based rendering has fascinating applications outside of traditional computer graphics. This collaboration between the RGL and LAPD groups at EPFL co-led by Baptiste Nicolet and Felix Wechsler demonstrates how to use it to create complex 3D prints very quickly.
We are excited to present a SIGGRAPH Asia paper exploring a new application of inverse rendering to Tomographic Volumetric Additive Manufacturing (TVAM), a new light-based 3D printing technology that can print objects in less than a minute.
November 27, 2024 at 2:23 PM
Differentiable physically-based rendering has fascinating applications outside of traditional computer graphics. This collaboration between the RGL and LAPD groups at EPFL co-led by Baptiste Nicolet and Felix Wechsler demonstrates how to use it to create complex 3D prints very quickly.
Reposted by Marco Salvi
Sharing a starter setup for using @shader-slang.bsky.social with #WebGPU in C++ with #CMake, either for native or web targets. 🔹 Uses the #reflection API to auto-generate boilerplate code! ✨
👉 github.com/eliemichel/S...
Very happy to take feedback about what's good and bad in this demo! More⤵️ 1/8
👉 github.com/eliemichel/S...
Very happy to take feedback about what's good and bad in this demo! More⤵️ 1/8
November 26, 2024 at 9:27 PM
Sharing a starter setup for using @shader-slang.bsky.social with #WebGPU in C++ with #CMake, either for native or web targets. 🔹 Uses the #reflection API to auto-generate boilerplate code! ✨
👉 github.com/eliemichel/S...
Very happy to take feedback about what's good and bad in this demo! More⤵️ 1/8
👉 github.com/eliemichel/S...
Very happy to take feedback about what's good and bad in this demo! More⤵️ 1/8
Reposted by Marco Salvi
Following over 1.5 years of hard work (w/@njroussel.bsky.social &@rtabbara.bsky.social), we just released a brand-new version of Dr.Jit (v1.0), my lab's differentiable rendering compiler along with an updated Mitsuba (v3.6). The list of changes is insanely long—here is what we're most excited about🧵
November 26, 2024 at 3:09 PM
Following over 1.5 years of hard work (w/@njroussel.bsky.social &@rtabbara.bsky.social), we just released a brand-new version of Dr.Jit (v1.0), my lab's differentiable rendering compiler along with an updated Mitsuba (v3.6). The list of changes is insanely long—here is what we're most excited about🧵
Reposted by Marco Salvi
GPU Zen 3 is out!
I want to draw attention to the 96(!) page article I was a small part of detailing all the work from tooth-to-tail we put into making ray tracing a reality in Cyberpunk 2077. It is exhaustive and should be a reference for anyone looking to make RTX a reality in their own engines.
I want to draw attention to the 96(!) page article I was a small part of detailing all the work from tooth-to-tail we put into making ray tracing a reality in Cyberpunk 2077. It is exhaustive and should be a reference for anyone looking to make RTX a reality in their own engines.
November 25, 2024 at 4:24 PM
GPU Zen 3 is out!
I want to draw attention to the 96(!) page article I was a small part of detailing all the work from tooth-to-tail we put into making ray tracing a reality in Cyberpunk 2077. It is exhaustive and should be a reference for anyone looking to make RTX a reality in their own engines.
I want to draw attention to the 96(!) page article I was a small part of detailing all the work from tooth-to-tail we put into making ray tracing a reality in Cyberpunk 2077. It is exhaustive and should be a reference for anyone looking to make RTX a reality in their own engines.
Reposted by Marco Salvi
My team RGL is looking for new PhD students starting 2025. If you are excited about topics like inverse rendering, compilers for graphics, and physically based modeling then please join us at EPFL. Info about the lab & admission process: rgl.epfl.ch/pages/jobs (Deadline: Dec 15)
November 22, 2024 at 3:14 PM
My team RGL is looking for new PhD students starting 2025. If you are excited about topics like inverse rendering, compilers for graphics, and physically based modeling then please join us at EPFL. Info about the lab & admission process: rgl.epfl.ch/pages/jobs (Deadline: Dec 15)
Reposted by Marco Salvi
Slang is now officially under open governance at Khronos, and what better way to celebrate than running Slang shaders directly in your browser with WebGPU?
try.shader-slang.org
You can also compile your shaders to a variety of targets including Metal and HLSL. I'm giddy to finally share this!
try.shader-slang.org
You can also compile your shaders to a variety of targets including Metal and HLSL. I'm giddy to finally share this!
Slang Playground
Interactive playground for experimenting with Slang shading language. Write, compile, and visualize shader code in real-time.
try.shader-slang.org
November 21, 2024 at 4:01 PM
Slang is now officially under open governance at Khronos, and what better way to celebrate than running Slang shaders directly in your browser with WebGPU?
try.shader-slang.org
You can also compile your shaders to a variety of targets including Metal and HLSL. I'm giddy to finally share this!
try.shader-slang.org
You can also compile your shaders to a variety of targets including Metal and HLSL. I'm giddy to finally share this!
Reposted by Marco Salvi
Khronos will oversee the open-source Slang shading language and compiler www.khronos.org/news/press/k...
This is big. Slang is cross-platform, backward-compatible, and trivial to port to. It supports differentiation/gradient computation as the first-class citizen, allowing to merge ML and graphics.
This is big. Slang is cross-platform, backward-compatible, and trivial to port to. It supports differentiation/gradient computation as the first-class citizen, allowing to merge ML and graphics.
Khronos Group Launches Slang Initiative, Hosting Open Source Compiler Contributed by NVIDIA
The Khronos Group has announced the launch of the new Slang™…
www.khronos.org
November 21, 2024 at 4:19 PM
Khronos will oversee the open-source Slang shading language and compiler www.khronos.org/news/press/k...
This is big. Slang is cross-platform, backward-compatible, and trivial to port to. It supports differentiation/gradient computation as the first-class citizen, allowing to merge ML and graphics.
This is big. Slang is cross-platform, backward-compatible, and trivial to port to. It supports differentiation/gradient computation as the first-class citizen, allowing to merge ML and graphics.
Reposted by Marco Salvi
SIGGRAPH Asia in two weeks, first paper thread on 🦋!
🎓 1/3 Controlmat, a diffusion model for material acquisition.
gvecchio.com/controlmat/
Given an image we generate a corresponding tileable, high-resolution relightable material!
🎓 1/3 Controlmat, a diffusion model for material acquisition.
gvecchio.com/controlmat/
Given an image we generate a corresponding tileable, high-resolution relightable material!
November 20, 2024 at 4:43 PM
SIGGRAPH Asia in two weeks, first paper thread on 🦋!
🎓 1/3 Controlmat, a diffusion model for material acquisition.
gvecchio.com/controlmat/
Given an image we generate a corresponding tileable, high-resolution relightable material!
🎓 1/3 Controlmat, a diffusion model for material acquisition.
gvecchio.com/controlmat/
Given an image we generate a corresponding tileable, high-resolution relightable material!
Reposted by Marco Salvi
Created a starter pack for researchers working in inverse graphics, 3D vision, and geometry processing.
Would love your help to expand this list!
go.bsky.app/9uEdjzb
Would love your help to expand this list!
go.bsky.app/9uEdjzb
November 18, 2024 at 2:43 PM
Created a starter pack for researchers working in inverse graphics, 3D vision, and geometry processing.
Would love your help to expand this list!
go.bsky.app/9uEdjzb
Would love your help to expand this list!
go.bsky.app/9uEdjzb
Is there a way to bookmark posts?
November 20, 2024 at 2:37 AM
Is there a way to bookmark posts?
Reposted by Marco Salvi
It would be amazing to see the computer graphics community engage on Bluesky, so here's a starter pack for those who might be interested in joining in:
go.bsky.app/K8ZL1kq
go.bsky.app/K8ZL1kq
November 20, 2024 at 12:54 AM
It would be amazing to see the computer graphics community engage on Bluesky, so here's a starter pack for those who might be interested in joining in:
go.bsky.app/K8ZL1kq
go.bsky.app/K8ZL1kq
Reposted by Marco Salvi
I've found a box of copies of my Deep Learning book nostarch.com/deep-learnin.... I'll ship one to the first 12 people who ask (via DM) & are willing to cover $11 postage. Rules: 1) US only (sorry, it's an expensive hassle otherwise) 2) You PayPal me the $ before I ship- I'll DM back w/ the details
November 19, 2024 at 10:53 PM
I've found a box of copies of my Deep Learning book nostarch.com/deep-learnin.... I'll ship one to the first 12 people who ask (via DM) & are willing to cover $11 postage. Rules: 1) US only (sorry, it's an expensive hassle otherwise) 2) You PayPal me the $ before I ship- I'll DM back w/ the details
It seems a lot of graphics and AI folks are moving over to Bluesky.
Since Threads has unfortunately, become just another doom scrolling machine, I hope this is the time we can ditch X for good.
Since Threads has unfortunately, become just another doom scrolling machine, I hope this is the time we can ditch X for good.
November 19, 2024 at 11:28 PM
It seems a lot of graphics and AI folks are moving over to Bluesky.
Since Threads has unfortunately, become just another doom scrolling machine, I hope this is the time we can ditch X for good.
Since Threads has unfortunately, become just another doom scrolling machine, I hope this is the time we can ditch X for good.