Collov AI’s Post

🚀 Introducing Stable Diffusion 3.5 by Stability AI 🎨 Unveiling the latest evolution in the Stable Diffusion family: Stable Diffusion 3.5, featuring new model variants designed for speed, quality, and accessibility. ✨ Stable Diffusion 3.5 Large - 8 billion parameters - Delivers superior image quality and prompt accuracy - most powerful model yet! ⚡ Stable Diffusion 3.5 Large Turbo - Distilled version of Large, with 4-step image generation - Same quality, but much faster! 🖼️ Stable Diffusion 3.5 Medium (Coming Oct 29): - 2.5 billion parameters - Optimized for consumer hardware - Balances quality and performance with resolutions up to 2 megapixels. Architecture: All models are built on Multimodal Diffusion Transformer (MMDiT) architecture, ensuring cutting-edge image generation with enhanced typography, prompt understanding, and resource efficiency. Muhammad Rizwan Munawar isn't it amazing, image generation models are more advanced now.  #AI #StableDiffusion #GenerativeAI #TechInnovation

🚨SIGGRAPH Asia 2024 Paper Alert 🚨 ➡️Paper Title: URAvatar: Universal Relightable Gaussian Codec Avatars 🌟Few pointers from the paper 🎯Authors of this paper presented a new approach to creating photorealistic and relightable head avatars from a phone scan with unknown illumination. 🎯The reconstructed avatars can be animated and relit in real time with the global illumination of diverse environments. 🎯Unlike existing approaches that estimate parametric reflectance parameters via inverse rendering, their approach directly models learnable radiance transfer that incorporates global light transport in an efficient manner for real-time rendering. 🎯However, learning such a complex light transport that can generalize across identities is non-trivial. A phone scan in a single environment lacks sufficient information to infer how the head would appear in general environments. 🎯 To address this, they built a universal relightable avatar model represented by 3D Gaussians. They trained on hundreds of high-quality multi-view human scans with controllable point lights. 🎯High-resolution geometric guidance further enhances the reconstruction accuracy and generalization. 🎯Once trained, they finetuned the pretrained model on a phone scan using inverse rendering to obtain a personalized relightable avatar. 🎯Their experiments establish the efficacy of their design, outperforming existing approaches while retaining real-time rendering capability. 🏢Organization: Codec Avatars Lab, Meta 🧙Paper Authors: Junxuan Li, Chen Cao, Gabriel Schwartz , Rawal Khirodkar, Christian Richardt, Tomas Simon, Yaser Sheikh, Shunsuke S. 📝  Read the Full Paper here: https://lnkd.in/g6TMAnEy 🗂️ Project Page: https://lnkd.in/gADCtJQe 🎥 Be sure to watch the attached Demo Video - Sound on 🔊🔊 Find this Valuable 💎 ? ♻️REPOST and teach your network something new Follow me 👣, Naveen Manwani, for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. #SIGGRAPHAsia2024

🔥 𝐏𝐫𝐞𝐬𝐞𝐧𝐭𝐢𝐧𝐠 “𝐕𝐢𝐝𝐮𝟒𝐃”: 𝐒𝐢𝐧𝐠𝐥𝐞 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐞𝐝 𝐕𝐢𝐝𝐞𝐨 𝐭𝐨 𝐇𝐢𝐠𝐡-𝐅𝐢𝐝𝐞𝐥𝐢𝐭𝐲 𝟒𝐃 𝐑𝐞𝐜𝐨𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐃𝐲𝐧𝐚𝐦𝐢𝐜 𝐆𝐚𝐮𝐬𝐬𝐢𝐚𝐧 𝐒𝐮𝐫𝐟𝐞𝐥𝐬!🔥 𝐌𝐨𝐭𝐢𝐯𝐚𝐭𝐢𝐨𝐧: Creating high-fidelity 4D reconstructions from single videos is a significant challenge, especially with non-rigid movements and frame distortions. The innovative teams from 𝐓𝐬𝐢𝐧𝐠𝐡𝐮𝐚 𝐔𝐧𝐢𝐯𝐞𝐫𝐬𝐢𝐭𝐲, 𝐒𝐡𝐞𝐧𝐠𝐒𝐡𝐮 𝐚𝐧𝐝 𝐓𝐨𝐧𝐠𝐣𝐢 𝐔𝐧𝐢𝐯𝐞𝐫𝐬𝐢𝐭𝐲 developed "𝐕𝐢𝐝𝐮4𝐃," a model that excels in capturing 4D representations using Dynamic Gaussian Surfels (DGS). This technique ensures precise motion depiction and texture details, enhancing both spatial and temporal coherence in virtual content creation. 👉 𝐄𝐱𝐩𝐥𝐨𝐫𝐞 𝐭𝐡𝐞 𝐅𝐮𝐥𝐥 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐚𝐧𝐝 𝐂𝐨𝐝𝐞 𝐡𝐞𝐫𝐞: https://lnkd.in/gTt7X9NE 𝐊𝐞𝐲 𝐅𝐞𝐚𝐭𝐮𝐫𝐞𝐬: ✅ 𝐃𝐲𝐧𝐚𝐦𝐢𝐜 𝐆𝐚𝐮𝐬𝐬𝐢𝐚𝐧 𝐒𝐮𝐫𝐟𝐞𝐥𝐬: Vidu4D employs DGS to dynamically warp surface elements, capturing detailed motion and deformation. ✅ 𝐖𝐚𝐫𝐩𝐢𝐧𝐠 𝐅𝐢𝐞𝐥𝐝𝐬: Optimized warping functions ensure the structural integrity of the 4D reconstructions. ✅ 𝐓𝐞𝐱𝐭𝐮𝐫𝐞 𝐀𝐜𝐜𝐮𝐫𝐚𝐜𝐲: Refined rotation and scaling parameters reduce texture flickering, capturing fine-grained details effectively. #Vidu4D #4DReconstruction #AI #Innovation #Technology 💚 Elevate your 4D reconstruction capabilities with Vidu4D – merging precision and innovation in virtual content creation! 💚 Follow me Matt Dinh to stay up to date with AI.

🔗 Excited to share this latest research in the field of architecture and artificial intelligence! Dive into a comprehensive literature review exploring the applications of generative AI in architectural design, from 2D images to 3D models. This paper highlights the impact of generative AI on architectural creativity and efficiency. Check it out here: https://bit.ly/3PNCPZH #AI #architecture #designinnovation

“If you can visualize it, you can build it.” Hassan Ragab. Join us this weekend for an exciting workshop with the architect, interdisciplinary designer, and conceptual artist Hassan Ragab on July 6-7, 2024. Learn how to use powerful AI tools like Midjourney, Stable Diffusion, and Krea AI to bring your design ideas to life. Register Now: Tap the🔗 link: https://lnkd.in/evYKJ5vs The workshop aims to immerse participants in the innovative use of AI text-to-image tools for architectural design. Participants will learn to craft effective prompts, develop intricate designs, and collaborate on groundbreaking concepts, enhancing their ability to integrate AI into their creative processes. 📑 Topic: AI Conceptual Architecture 4.0 📅 Date: July 6 - 7, 2024 🕕 Time: 16:00 - 20:00 GMT ⚒️ Software: Midjourney, Stable Diffusion and ControlNet, ComfyUI, KREA, Magnific Ai, RunwayML, Automatic1111 🧑🏼🎓 Total Seats: 50 Seats 🛒 15% Discount for Digital Members. 🏷 Three workshops are offered in Artificial Intelligence Bundle 4.0 with a 25% discount and an additional %15 discount for digital members: https://lnkd.in/e_RframT #artificialintelligence #midjourney #parametricdesign #computationaldesign #architecturestudents #formgeneration

The Dream Machine - 1607 #AIArt Prompt: Generate a digital art image that depicts a technology-inspired environment. The atmosphere should radiate joy and happiness, perhaps fostering a sense of excitement about the potential of technology. The quality of the illustration should be simplified - focusing more on the shapes, colors, and objects rather than intricate details. Perhaps, it can showcase figures engaged with various technological devices, set against a backdrop of a futuristic city or a hi-tech room. The use of vibrant colors and simple, clean lines would help encapsulate the mood and the theme.

🚨CVPR 2024 Paper Alert 🚨 ➡️Paper Title: DiffHuman: Probabilistic Photorealistic 3D Reconstruction of Humans 🌟Few pointers from the paper 🎯In this paper authors have presented “DiffHuman”, a probabilistic method for photorealistic 3D human reconstruction from a single RGB image. Despite the ill-posed nature of this problem, most methods are deterministic and output a single solution, often resulting in a lack of geometric detail and blurriness in unseen or uncertain regions. 🎯In contrast, DiffHuman predicts a probability distribution over 3D reconstructions conditioned on an input 2D image, which allowed them to sample multiple detailed 3D avatars that are consistent with the image. DiffHuman is implemented as a conditional diffusion model that denoises pixel-aligned 2D observations of an underlying 3D shape representation. 🎯 During inference, authors may sample 3D avatars by iteratively denoising 2D renders of the predicted 3D representation. Furthermore, authors have also introduced a generator neural network that approximates rendering with considerably reduced runtime (55x speed up), resulting in a novel dual-branch diffusion framework. 🎯 Their experiments showed that DiffHuman can produce diverse and detailed reconstructions for the parts of the person that are unseen or uncertain in the input image, while remaining competitive with the state-of-the-art when reconstructing visible surfaces. 🏢Organization:  Google Research, University of Cambridge 🧙Paper Authors: Akash Sengupta, Thiemo Alldieck, Nikos Kolotouros, Enric Corona, Andrei Zanfir, Cristian Sminchisescu 1️⃣Read the Full Paper here: https://lnkd.in/gYTe6-R8 2️⃣Project Page: https://lnkd.in/gzKRpCjt 🎥 Be sure to watch the attached Informational Video-Sound on 🔊🔊 🎵 Music by Dmytro Kuvalin from Pixabay Find this Valuable 💎 ? ♻️REPOST and teach your network something new Follow me 👣, Naveen Manwani, for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. #CVPR2024

AI and generative design can improve quality, reduce waste, and lower costs across the entire A/ E/C industry. They reduce the risks of design and fabrication and provide a platform that gives reliable results during a structure's design phase. See how this works in this interesting @HP article outlining the advantages

The field of structural engineering is on the brink of a transformative revolution, driven by advancements in artificial intelligence. AI-powered generative design algorithms in structural engineering create optimized structural designs by: - Exploring numerous configurations - Improving material efficiency - Reducing costs - Enhancing structural performance under various environmental conditions We at BERM believe that integration of new technology is essential for staying relevant in the structural world. #AI #GenerativeDesign #StructuralEngineering #MaterialEfficiency #CostReduction #PerformanceEnhancement #InnovativeEngineering #SustainableDesign

We tested SF3D by Stability AI and here the result. They released a 3D asset generator from images, complete with textures, UV mapping, and materials. The generation process takes just a few seconds on the H100. We conducted a comparative test with DreamCraft3D, which our artists are currently using in production. While the speed is certainly impressive, it comes at the cost of lower quality. In the comparison of the Stable Fast 3D Mesh Reconstruction (on the right) and DreamCraft3D (on the left). DreamCraft3D has more detailed textures and mesh. Also Stable Fast 3D Mesh Reconstruction has some gray artifacts and general shape in incorrect. But It's worth noting that DreamCraft3D generation takes 2 machine hours and Stable Fast 3D Mesh Reconstruction generation takes 5 seconds on a similar hardware. #AI #3DRendering #Imageto3D #3DModeling #TechNews #StabilityAI #Innovation