Edward Y. Chang’s Post

I am super excited to share that all the Chapters of my #book "#GenAI in Action" 📔 are now available as part of Manning Publications Co. early access program. You can read more here: https://lnkd.in/gdGU5T3t The book is going through final reviews and hopefully off to printing soon. 🙃 #AI #OpenAI #AzureAI #GenerativeAI

Prior to Gutenberg’s invention of the movable-type printing press around 1440, the church controlled information, including the classical writings of Plato, Socrates, Thucydides, et al. What happened following Gutenberg? Information was released into the hands of society, powering the Renaissance, the Reformation, the Enlightenment, and the first nation state based on Enlightenment thinking: the US republic. GenAI is the Gutenberg press of the 21st century. It will unlock information and create the means for new and original data to be created. #genai #AI #artificialintelligence

From text-to-text to... "speech-to-avatar"!!! 🤖 Yesterday I was talking with the VP of Engineering of a company dealing with AI Avatars and Digital Twins and for the first time I've heard this wording. Speech-to-avatar It really made me think and then, surprise surprise, I got caught into this article by John Nosta, exploring the "cognitive interface" - a language-based, cognitively responsive UI design that adapts to the user's needs and context. "Gutenberg's printing press democratized access to words, while the Internet revolutionized the dissemination of facts. Now, LLMs are unlocking the realm of ideas. The cognitive interface enables users to engage with machines on a conceptual level, moving beyond the retrieval of information to the exploration and refinement of thoughts." We are all witnessing an amazing revolution. Welcome to the Cognitive Age. #RegulateAI

🚨Paper Alert 🚨 ➡️Paper Title: MarDini: Masked Autoregressive Diffusion for Video Generation at Scale 🌟Few pointers from the paper 🎯Authors of this paper introduced “MarDini”, a new family of video diffusion models that integrate the advantages of masked auto-regression (MAR) into a unified diffusion model (DM) framework. 🎯Here, MAR handles temporal planning, while DM focuses on spatial generation in an asymmetric network design: ➕ A MAR-based planning model containing most of the parameters generates planning signals for each masked frame using low-resolution input ➕ A lightweight generation model uses these signals to produce high-resolution frames via diffusion de-noising. 🎯MarDini's MAR enables video generation conditioned on any number of masked frames at any frame positions: 👀 A single model can handle video interpolation (e.g., masking middle frames) 👀 Image-to-video generation (e.g., masking from the second frame onward) 👀 Video expansion (e.g., masking half the frames) 🎯The efficient design allocates most of the computational resources to the low-resolution planning model, making computationally expensive but important spatio-temporal attention feasible at scale. 🎯MarDini sets a new state-of-the-art for video interpolation; meanwhile, within few inference steps, it efficiently generates videos on par with those of much more expensive advanced image-to-video models. 🏢Organization: AI at Meta , KAUST (King Abdullah University of Science and Technology) 🧙Paper Authors: Haozhe Liu, Shikun Liu, Zijian Zhou, Mengmeng Xu, Yanping Xie, Xiao Han, Juan C. Pérez, Ding Liu, Kumara Kahatapitiya, Menglin Jia, Jui-Chieh Wu, Sen He, Tao Xiang, Jürgen Schmidhuber, Juan-Manuel Pérez-Rúa 1️⃣Read the Full Paper here: https://lnkd.in/gjvmNj6D 2️⃣Project Page: https://lnkd.in/gBmbNnVs 🎥 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.

I'm sharing my latest blog post, which delves into the innovative research presented in "Generative Image Dynamics," a notable paper from CVPR 2024. This paper introduces a novel method for generating realistic motion from still images. Here’s a concise overview of the key topics discussed: ✨ Motion Representation: Exploration of "Spectral Volume," 🧠 a frequency-domain approach to capture long-term motion patterns. 🖼️ Motion Prediction: Examination of how a frequency-coordinated diffusion model predicts complex motion from static imagery. If you are interested in understanding how AI is being used to model natural motion and transform still images into dynamic visuals, I encourage you to read the full analysis. The complex methodologies are broken down into clear and digestible sections. I welcome your comments and feedback. What are your insights on this paper and the potential impact of this research on future visual content creation? Let’s discuss.

🚨Paper Alert 🚨 ➡️Paper Title: Reconstructing People, Places, and Cameras 🌟Few pointers from the paper 🎯Authors of this paper presented "Humans and Structure from Motion" (HSfM), a method for jointly reconstructing multiple human meshes, scene point clouds, and camera parameters in a metric world coordinate system from a sparse set of uncalibrated multi-view images featuring people. 🎯 Their approach combines data-driven scene reconstruction with the traditional Structure-from-Motion (SfM) framework to achieve more accurate scene reconstruction and camera estimation, while simultaneously recovering human meshes. 🎯 In contrast to existing scene reconstruction and SfM methods that lack metric scale information, their method estimates approximate metric scale by leveraging a human statistical model. 🎯Furthermore, it reconstructs multiple human meshes within the same world coordinate system alongside the scene point cloud, effectively capturing spatial relationships among individuals and their positions in the environment. 🎯They initialized the reconstruction of humans, scenes, and cameras using robust foundational models and jointly optimize these elements. 🎯This joint optimization synergistically improves the accuracy of each component. 🎯They compared their method to existing approaches on two challenging benchmarks, EgoHumans and EgoExo4D, demonstrating significant improvements in human localization accuracy within the world coordinate frame (reducing error from 3.51m to 1.04m in EgoHumans and from 2.9m to 0.56m in EgoExo4D). 🎯Notably, their results show that incorporating human data into the SfM pipeline improves camera pose estimation (e.g., increasing RRA@15 by 20.3% on EgoHumans). 🎯Additionally, qualitative results show that their approach improves overall scene reconstruction quality. 🏢Organization:   University of California, Berkeley 🧙Paper Authors:  Lea Müller, Hongsuk Choi, Anthony Zhang, Brent Yi, Jitendra MALIK, Angjoo Kanazawa 📝 Read the Full Paper here: https://lnkd.in/gFbAupS3 🗂️ Project Page: https://lnkd.in/gtRFe432    🎥 Be sure to watch the attached Demo Video - Sound on 🔊🔊 🎵 Music by Restum Anoush 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.

This is such a huge step-up from PHOSA framework for human-object interaction 3D mesh reconstruction 👏

𝐋𝐋𝐌2𝐅𝐄𝐀: 𝐃𝐢𝐬𝐜𝐨𝐯𝐞𝐫 𝐍𝐨𝐯𝐞𝐥 𝐃𝐞𝐬𝐢𝐠𝐧𝐬 𝐰𝐢𝐭𝐡 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐄𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐚𝐫𝐲 𝐌𝐮𝐥𝐭𝐢𝐭𝐚𝐬𝐤𝐢𝐧𝐠 📘 𝐖𝐡𝐚𝐭 𝐢𝐬 𝐭𝐡𝐢𝐬 𝐩𝐚𝐩𝐞𝐫 𝐚𝐛𝐨𝐮𝐭?  This paper introduces LLM2FEA, a novel framework that combines Large Language Models (LLMs) with evolutionary multitasking to discover innovative designs. The framework leverages the strengths of generative models and evolutionary algorithms to enhance design efficiency and creativity. 🤖 𝐊𝐞𝐲 𝐂𝐨𝐦𝐩𝐨𝐧𝐞𝐧𝐭𝐬: Generative Models: Utilizes LLMs to generate initial design concepts. Evolutionary Algorithms: Applies evolutionary strategies to refine and optimize these designs. 📊 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞𝐬: Enhance design creativity by combining AI-driven generative models with evolutionary algorithms. 🧠 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐢𝐯𝐞 𝐀𝐩𝐩𝐫𝐨𝐚𝐜𝐡: Merges the capabilities of LLMs with evolutionary algorithms, a novel combination in the field of design and optimization. Employs a multitasking strategy to leverage synergies between different tasks, enhancing overall performance. 🚀 𝐖𝐡𝐲 𝐢𝐬 𝐭𝐡𝐢𝐬 𝐚 𝐛𝐫𝐞𝐚𝐤𝐭𝐡𝐫𝐨𝐮𝐠𝐡? Efficiency: Significantly reduces the time required to arrive at optimal design solutions. Creativity: Unlocks new possibilities in design by harnessing the generative power of LLMs. Scalability: Applicable to a wide range of design challenges across various industries. ⏱ 𝐊𝐞𝐲 𝐅𝐢𝐧𝐝𝐢𝐧𝐠𝐬: Performance Improvement: Demonstrates superior performance in generating and optimizing designs compared to traditional methods. Innovation Potential: Shows great potential in discovering novel design solutions that were previously unattainable. Task Synergy: Effectively utilizes multitasking to enhance the de

Building an efficient AI comic creation platform means we need to make every second count ⏱️ To deliver rapid feedback during multiple image generation iterations, we turned to Adversarial Diffusion Distillation — a technique that speeds up image creation without sacrificing quality. In our latest post on the Dashtoon Insiders blog, Ayushman Buragohain shares: 🎯 Why we chose this technique  🎯 How we implemented it with in-house models  🎯 Our steps to streamline the process This has been a game-changer for both speed and creativity at Dashtoon Studio Read the full blogpost at https://lnkd.in/g2EeazZU

𝐋𝐋𝐌2𝐅𝐄𝐀: 𝐃𝐢𝐬𝐜𝐨𝐯𝐞𝐫 𝐍𝐨𝐯𝐞𝐥 𝐃𝐞𝐬𝐢𝐠𝐧𝐬 𝐰𝐢𝐭𝐡 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐄𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐚𝐫𝐲 𝐌𝐮𝐥𝐭𝐢𝐭𝐚𝐬𝐤𝐢𝐧𝐠 📘 𝐖𝐡𝐚𝐭 𝐢𝐬 𝐭𝐡𝐢𝐬 𝐩𝐚𝐩𝐞𝐫 𝐚𝐛𝐨𝐮𝐭? This paper introduces LLM2FEA, a novel framework that combines Large Language Models (LLMs) with evolutionary multitasking to discover innovative designs. The framework leverages the strengths of generative models and evolutionary algorithms to enhance design efficiency and creativity. 🤖 𝐊𝐞𝐲 𝐂𝐨𝐦𝐩𝐨𝐧𝐞𝐧𝐭𝐬: Generative Models: Utilizes LLMs to generate initial design concepts. Evolutionary Algorithms: Applies evolutionary strategies to refine and optimize these designs. 📊 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞𝐬: Enhance design creativity by combining AI-driven generative models with evolutionary algorithms. 🧠 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐢𝐯𝐞 𝐀𝐩𝐩𝐫𝐨𝐚𝐜𝐡: Merges the capabilities of LLMs with evolutionary algorithms, a novel combination in the field of design and optimization. Employs a multitasking strategy to leverage synergies between different tasks, enhancing overall performance. 🚀 𝐖𝐡𝐲 𝐢𝐬 𝐭𝐡𝐢𝐬 𝐚 𝐛𝐫𝐞𝐚𝐤𝐭𝐡𝐫𝐨𝐮𝐠𝐡? Efficiency: Significantly reduces the time required to arrive at optimal design solutions. Creativity: Unlocks new possibilities in design by harnessing the generative power of LLMs. Scalability: Applicable to a wide range of design challenges across various industries. ⏱ 𝐊𝐞𝐲 𝐅𝐢𝐧𝐝𝐢𝐧𝐠𝐬: Performance Improvement: Demonstrates superior performance in generating and optimizing designs compared to traditional methods. Innovation Potential: Shows great potential in discovering novel design solutions that were previously unattainable. Task Synergy: Effectively utilizes multitasking to enhance the de