Gradio

?? New Research: Self-training inspires clear and concise thinking in LLMs! 30% reduction in tokens in LLM output, across five model families on GSM8K and MATH while maintaining average accuracy ?? Useful links for Concise-Reasoning: > Code: https://lnkd.in/gPaeuXKE > Official app on Hugging Face : https://lnkd.in/gyG_hsri

DarkIR ?????? Now you can take photographs at night or in low-light conditions without concern. Say goodbye to noise, low light, and blurriness problems. Learn more about the new research or build locally: https://lnkd.in/gKEbuiC7 DarkIR app is live on Hugging Face: https://lnkd.in/gAXmPef8

????This is amazing! You can now create high-quality 3D Scenes from a single image using MIDI (Multi-Instance Diffusion Models) > Explore more about the MIDI Project: https://lnkd.in/gdBQqtWB > And keep an eye out as the Gradio app is being prepared here on Hugging Face : https://lnkd.in/gtzUzZhK

Text to generative 3D worlds, right on your desktop! ?? Follow me as I head NVIDIA GTC to go deep into this tech with the experts. To celebrate, I'm giving away a brand-new RTX 4080 SUPER to one lucky follower! This GPU can handle just about any 3DGS or AI project you want to throw at it...and entering my giveaway is easy! Here's how to enter: 1???Sign up for your favorite GTC session here (Yes, Jensen's keynote counts!): https://nvda.ws/4h8kz8X 2?? Fill out the form linked below with a screenshot or photo from a session and 1-2 sentences of your favorite takeaway from the session. Link to enter the giveaway: https://lnkd.in/g2yB88UF If you are planning on being at the event in person, DM me! I would love to connect! About the tech stack behind the video: I used NVIDIA Cosmos using my custom Gradio interface > Text2World gen output trained in jawset visual computing's PostShot into gaussian splats where I also rendered a camera flythrough. Jessica Ji Michael Rubloff Michal Gula Bilawal Sidhu #3D #AI #GenerativeAI #GTC25

Just wrapped up porting mast3r-slam to Rerun and integrating a Gradio interface—exciting to see it work on most videos! Here are the highlights: ? **Performance:** On my 3060, it runs at about 3 fps. Not real-time yet, but stronger hardware could change that. ? **Optimization Insights:** ? Backend factor graph optimization significantly improves camera extrinsics but has little effect on geometry consistency. ? Reducing resolution from 512 to 224 nearly doubles speed, though it compromises accuracy, especially for dense geometry. ? **Nerfstudio Integration:** I’ve included a zip file with keyframes, camera poses, and a point cloud for use with Nerfstudio (with plans to add depth maps later). ? **Acknowledgements:** Thanks to Riku Murai, Eric Dexheimer, and Andrew Davison for open-sourcing this?project—achieving robust SLAM from a single video is really incredible

?? A New Sota App for Low-light Image Enhancement : HVI Improves the visual quality of images captured in low-light environments while suppressing noise and color distortion. Check out the official gradio app on Hugging Face ??: https://lnkd.in/g_57Uk7K OR, Build HVI locally with Gradio : https://lnkd.in/gpZCpcXh HVI, which stands for Horizontal Vertical Color Intensity, represents a new research initiative and tool aimed at enhancing low-light images. Explore this research further through their app and model available on Hugging Face.

Curious on how to build local voice-drivent AI assistants? Check out Chee Kin Lim's tutorial on FastRTC Write-up: https://lnkd.in/erWXVMc6 Code: https://lnkd.in/eVzK7Bfn

YOLOe: Real-Time Seeing Anything Just like the human eye, it is a unified object detection and segmentation model. Accepts different prompt styles--texts, visual inputs, and prompt-free generation. Model and App released on Hugging Face. Free App: https://lnkd.in/gGDird8G Run YOLOe locally with Gradio: https://lnkd.in/gesYvBTf

Next.js + FastRTC: Production Audio Streaming Made Simple ?? Check out this impressive demo by Rohan showcasing seamless integration between FastRTC and @nextjs! The video demonstrates: ? Beautiful UI components built with Next.js ? Production-grade WebRTC server powered by FastRTC ? How to integrate FastRTC with any UI framework Building production audio streaming applications in Python has never been easier. The complete code is available on GitHub - https://lnkd.in/egd4u6Va kudos Rohan Richard

Diffusion Self-Distillation (DSD) for zero-shot Image customization! ?? Get Fine-grained Control over the text-to-image diffusion process. What DSD can do for you ?? - > Character Preservation > Item Preservation > Instruction Prompting > Relighting > Comic Generation Official App is live on Hugging Face : https://lnkd.in/gUvX-XEW