More
    HomeAI Papers

    AI Papers

    One Token to Seg Them All: VideoLISA from Amazon for Language-Instructed Video Segmentation

    Approach to Object Segmentation in Videos Using Language Instructions Language-Instructed Reasoning: VideoLISA leverages the capabilities of large language models to create temporally consistent segmentation masks...

    Nvidia Shakes Up the AI Landscape: Meet NVLM 1.0, the Open-Source Giant Ready to Rival GPT-4

    A Revolutionary Move Towards Accessibility and Innovation in Artificial Intelligence Nvidia has made a significant splash in the AI arena with its latest announcement: the...

    Unmasking Replication: Introducing ICDiff for Detecting Copying in Diffusion Models

    A New Approach to Ensure Originality in AI-Generated Images Challenge of Content Replication: While diffusion models can create stunning images, they may inadvertently replicate existing...

    3DTOPIA-XL: Revolutionizing 3D Asset Generation with Advanced Diffusion Techniques

    New Model Addresses Industry Demands for High-Quality, Efficient 3D Content Creation Transformative Technology: 3DTOPIA-XL introduces a novel primitive-based 3D representation, PrimX, which enables the generation...

    Mastering the Strings: Synchronizing Dual Hands for Realistic Guitar Playing

    A groundbreaking approach enables virtual guitarists to play complex rhythms and chords with precision and naturalness. Researchers present a novel method for synthesizing dexterous hand...

    Unleashing Creativity: MaskBit Image Generation

    A New Era of Embedding-Free Generation Using Bit Tokens MaskBit introduces a groundbreaking approach to image generation by utilizing bit tokens instead of traditional embeddings. The...

    MaskedMimic from Nvidia: Transforming Character Control with Unified Motion Inpainting

    Revolutionizing Animation: A New Approach to Physics-Based Character Interaction MaskedMimic introduces a novel physics-based character control system that synthesizes motions from partial input descriptions. This unified...

    NASA’s SPAR Lab Unveils OnAIR: The Future of Autonomous Spacecraft AI

    Empowering Spacecraft with Intelligence: A Leap Toward Autonomous Exploration NASA’s SPAR Lab introduces OnAIR, an AI platform designed to enhance spacecraft resilience and autonomy. The platform...

    Imagine Yourself: Personalized Image Generation Without Tuning

    Breaking Barriers in Image Synthesis: How Meta's New Model Redefines Customization Meta introduces "Imagine Yourself," a cutting-edge model for personalized image generation that operates without...

    GRIN MoE: AI Efficiency with Microsoft’s Latest Model

    Unlocking the Future of Deep Learning: A Deep Dive into GRIN MoE Microsoft introduces GRIN MoE, an innovative Gradient-Informed Mixture of Experts model designed to...

    FlexiTex: Texture Generation with Visual Precision

    Harnessing Visual Guidance to Create High-Quality Textures for 3D Assets Bridging the Gap: Traditional texture generation methods often rely on abstract text prompts, leading to inconsistent...

    The Power of Sparse Computation from Microsoft: GRIN for Enhanced Mixture-of-Experts Training

    MoE Models with Gradient-Informed Techniques to Boost Performance and Scalability Advancements in Mixture-of-Experts (MoE) Models: MoE models leverage sparse computation to improve scalability, activating only a...