Ziwei Liu

New methods improve human pose and shape estimation from 3D data.

A new method improves wireless signals using Intelligent Reflecting Surfaces and deep reinforcement learning.

Introducing a framework for efficient 3D toonification while preserving original features.

A new method improves inbetweening for anime-style line drawings using technology.

Research on improving human pose estimation through diverse datasets and model scaling.

Vlogger simplifies video blogging, making it quicker and easier for creators.

A new framework enhances the prediction of protein interactions and functions.

A new system for converting text descriptions into detailed 3D models efficiently.

A method to estimate human and camera movements accurately using a single camera.

Research aims to improve machine interpretation of 3D environments for safety.

AiOS combines human pose detection and estimation into a single efficient process.

A new method improves text-to-image generation with smooth transitions and high quality.

A new method enhances 3D scene animations from text prompts.

A new method improves the realism and editability of 3D humans.

SurMo enhances video rendering of dynamic humans by merging appearance and motion.

MMInA evaluates how well agents perform tasks across multiple websites.

New models enhance duet interactions in virtual dance performances.

M3Net enhances LiDAR segmentation for self-driving cars by integrating diverse datasets and sensors.

Combining LiDAR and camera data to improve self-driving technology efficiency.

Teams worldwide enhance self-driving car technology in challenging conditions.

Introducing RoboBEV to test BEV algorithms under real-world conditions.

A detailed look at a dataset capturing everyday human activities.

SuperFlow enhances 3D perception models using LiDAR and camera data for autonomous driving.

Analyzing the importance and difficulties of assessing multimodal AI models.

ReSyncer improves video quality and flexibility for lip movements synchronized to audio.

Introducing TRTC systems to address challenges in mobile communication technology.

A new method improves AI performance across different image types.

Disco4D generates realistic 3D human figures from a single image.

Omni6D dataset enhances object pose estimation with diverse categories and realistic scenarios.

A new framework enhances motion tracking using wearables and egocentric videos.

BVTON technology brings lifelike virtual clothing fitting to online shoppers.

Learn how to create rich 3D textures from simple text descriptions.

Discover how lifelike characters transform virtual interactions.

Explore how Bullet Timer transforms videos into dynamic 3D scenes.