Meta tags:
author=Hongyu Liu;
description=A simple, whitespace theme for academics. Based on [*folio](htt????/github.com/bogoli/-folio) design. ;
keywords=jekyll, jekyll-theme, academic-website, portfolio-website;

Headings (most frequently used words):

recent, hongyu, liu, news, highlights, publications, dyaplex, longcat, video, avatar, opsd,

Text of the page (most frequently used words):
and (54), the (33), generation (17), for (13), video (12), avatar (11), 2026 (10), that (10), liu (9), with (9), #hongyu (8), from (8), vision (8), image (8), model (8), wang (7), computer (6), gaussian (6), pattern (5), xuan (5), wan (5), ziyu (5), chen (5), open (5), avatarization (5), head (5), autoregressive (5), interaction (5), conference (4), recognition (4), cvpr (4), domain (4), this (4), training (4), high (4), avatars (4), more (4), self (4), distillation (4), intelligence (4), avatarpointillist (4), point (4), long (4), technical (4), report (4), research (4), 2025 (3), ieee (3), yue (3), shen (3), yujun (3), qifeng (3), avatarartist (3), generative (3), diffusion (3), our (3), data (3), prior (3), its (3), across (3), quality (3), source (3), code (3), abs (3), authors (3), language (3), controlnet (3), realistic (3), editing (3), decoder (3), publications (3), step (3), longcat (3), speech (3), motion (3), digital (3), released (3), humans (3), spatial (3), github (2), have (2), contain (2), preview_fit (2), year (2), proceedings (2), cvf (2), journal (2), jingye (2), song (2), yibing (2), author (2), title (2), article (2), portrait (2), triplanes (2), practical (2), paradigm (2), gans (2), models (2), images (2), supervision (2), various (2), domains (2), between (2), construction (2), extensive (2), experiments (2), producing (2), pdf (2), html (2), bib (2), headartist (2), method (2), guided (2), score (2), ssd (2), pipeline (2), under (2), where (2), sampled (2), pose (2), guide (2), rendering (2), reference (2), novel (2), view (2), zijian (2), yating (2), generating (2), dynamic (2), points (2), also (2), fidelity (2), controllable (2), recent (2), policy (2), post (2), few (2), generators (2), opsd (2), audio (2), driven (2), full (2), duplex (2), dyadic (2), dyaplex (2), nvidia (2), accepted (2), may (2), jun (2), about (2), interested (2), systems (2), graphics (2), coherent (2), form (2), interactive (2), real (2), time (2), human (2), centered (2), multimodal (2), closed (2), loop (2), science (2), current (2), copyright, powered, theme, hosted, photos, unsplash, pages, folio, jekyll, you, any, questions, please, contact, fan, yanbo, liu2025avatarartist, work, focuses, purpose, creating, arbitrary, style, select, parametric, intermediate, representation, propose, takes, advantage, both, adversarial, networks, design, stems, observation, excel, bridging, without, yet, usually, face, challenges, handling, diverse, distributions, robust, emerges, solution, assisting, gan, transferring, expertise, synergy, these, experts, permits, multi, triplane, dataset, which, drives, development, general, creator, suggest, capable, strong, robustness, will, made, publicly, available, facilitate, future, studies, present, supports, either, input, utilizing, landmark, introducing, optimizes, parameterized, itself, given, camera, landmarks, are, rendered, noised, processed, through, frozen, twice, different, classifier, free, guidance, cfg, weights, alignment, prompts, encoder, extracts, identity, embeddings, sent, while, synthesizes, sculptures, rich, geometry, photo, appearance, support, operations, significantly, outperforming, state, art, methods, transactions, analysis, machine, tpami, runtao, zhou, boyao, liu2026avatarpointillist, introduce, framework, single, core, only, transformer, autoregressively, generates, cloud, splatting, sequential, approach, allows, precise, adaptive, dynamically, adjusting, density, total, number, based, subject, complexity, during, jointly, predicts, per, binding, information, enabling, animation, after, dedicated, converts, into, complete, renderable, attributes, demonstrate, conditioning, latent, features, generator, enables, effective, stages, markedly, improves, validate, produces, photorealistic, believe, formulation, represents, new, reducing, horizon, degradation, meigen, stronger, lip, sync, stability, inference, meituan, coupling, streaming, body, responsive, agents, highlights, siggraph, easyvfx, system, eccv, lacon, text, uncurated, project, jul, news, experience, affiliations, beyond, deeply, passionate, fishing, hope, one, day, travel, china, fish, rivers, lakes, coastal, waters, using, journey, reconnect, nature, explore, local, cultures, find, ideas, outside, lab, particularly, building, intelligent, bridge, recently, been, increasingly, especially, toward, expressive, temporally, centers, algorithms, specifically, pursue, agenda, progresses, further, forms, layered, understanding, reconstructing, animating, dimensions, spatiotemporal, static, reconstruction, student, department, engineering, hong, kong, university, technology, fortunate, advised, prof, hkust, spanning, focus, google, scholar, neural, content, phd, candidate, memories, projects, toggle, navigation,

Text of the page (random words):
hongyu liu toggle navigation about current publications projects memories hongyu liu phd candidate computer vision graphics digital humans 2d 3d content generation neural rendering video generation avatar intelligence view publications google scholar github current focus human centered multimodal spatial intelligence and generation real time video generation and long form interactive avatar video generation a closed research loop spanning 2d editing 3d generation and 4d interaction i am a ph d student in the department of computer science and engineering at the hong kong university of science and technology hkust where i am fortunate to be advised by prof qifeng chen my research centers on human centered multimodal spatial intelligence and generative algorithms more specifically i pursue a coherent research agenda that progresses from 2d spatial reconstruction and editing to 3d high fidelity static generation and further to 4d dynamic spatiotemporal interaction this forms a layered and closed loop pipeline for understanding reconstructing generating and animating realistic humans and avatars across dimensions i am particularly interested in building intelligent systems that bridge vision graphics and generation for digital humans recently i have also been increasingly interested in real time video generation and long form interactive avatar video generation especially toward controllable expressive and temporally coherent avatar systems beyond research i am deeply passionate about fishing i hope to one day travel across china to fish in its rivers lakes and coastal waters using the journey to reconnect with nature explore local cultures and find ideas outside the lab experience affiliations news jul 09 2026 we released opsd v an on policy self distillation project for post training few step autoregressive video generators jun 30 2026 lacon training text to image model from uncurated data is accepted to eccv 2026 jun 02 2026 we released dyaplex nvidia s full duplex speech motion model for dyadic interaction may 26 2026 we released the longcat video avatar 1 5 technical report an open source audio driven avatar video generation system may 22 2026 easyvfx is accepted to siggraph 2026 recent highlights nvidia technical report dyaplex full duplex speech motion model for dyadic interaction coupling streaming speech and body motion for responsive digital agents meituan longcat technical report longcat video avatar 1 5 open source audio driven avatar video generation with stronger lip sync long video stability and practical 8 step inference meigen technical report opsd v on policy self distillation for post training few step autoregressive video generators reducing long horizon degradation recent publications avatarpointillist autoregressive 4d gaussian avatarization hongyu liu xuan wang zijian wu yating wang ziyu wan and 5 more authors conference on computer vision and pattern recognition cvpr 2026 abs bib html pdf code we introduce avatarpointillist a novel framework for generating dynamic 4d gaussian avatars from a single portrait image at the core of our method is a decoder only transformer that autoregressively generates a point cloud for 3d gaussian splatting this sequential approach allows for precise adaptive construction dynamically adjusting point density and the total number of points based on the subject s complexity during point generation the ar model also jointly predicts per point binding information enabling realistic animation after generation a dedicated gaussian decoder converts the points into complete renderable gaussian attributes we demonstrate that conditioning the decoder on the latent features from the ar generator enables effective interaction between stages and markedly improves fidelity extensive experiments validate that avatarpointillist produces high quality photorealistic and controllable avatars we believe this autoregressive formulation represents a new paradigm for avatar generation article liu2026avatarpointillist title avatarpointillist autoregressive 4d gaussian avatarization author liu hongyu and wang xuan and wu zijian and wang yating and wan ziyu and ma yue and liu runtao and zhou boyao and shen yujun and chen qifeng journal conference on computer vision and pattern recognition cvpr year 2026 preview_fit contain headartist vl vision language guided 3d head generation with self score distillation hongyu liu xuan wang ziyu wan yujun shen yibing song and 2 more authors ieee transactions on pattern analysis and machine intelligence tpami 2025 abs we present headartist vl a 3d head generation method that supports either vision or language input utilizing a landmark guided controlnet as a generative prior and introducing a self score distillation ssd pipeline that optimizes a parameterized 3d head model under the supervision of the prior itself where given a sampled camera pose an image and its landmarks are rendered from the head model noised and processed through a frozen controlnet twice with different classifier free guidance cfg weights to guide rendering alignment with language prompts or for reference images an image encoder extracts identity embeddings sent to the controlnet while a novel view diffusion model synthesizes the reference image under the sampled pose to guide ssd producing high quality 3d head sculptures with rich geometry photo realistic appearance and support for editing operations significantly outperforming state of the art methods avatarartist open domain 4d avatarization hongyu liu xuan wang ziyu wan yue ma jingye chen and 4 more authors proceedings of the ieee cvf conference on computer vision and pattern recognition cvpr 2025 abs bib html pdf code this work focuses on open domain 4d avatarization with the purpose of creating a 4d avatar from a portrait image in an arbitrary style we select parametric triplanes as the intermediate 4d representation and propose a practical training paradigm that takes advantage of both generative adversarial networks gans and diffusion models our design stems from the observation that 4d gans excel at bridging images and triplanes without supervision yet usually face challenges in handling diverse data distributions a robust 2d diffusion prior emerges as the solution assisting the gan in transferring its expertise across various domains the synergy between these experts permits the construction of a multi domain image triplane dataset which drives the development of a general 4d avatar creator extensive experiments suggest that our model avatarartist is capable of producing high quality 4d avatars with strong robustness to various source image domains the code the data and the models will be made publicly available to facilitate future studies article liu2025avatarartist title avatarartist open domain 4d avatarization author liu hongyu and wang xuan and wan ziyu and ma yue and chen jingye and fan yanbo and shen yujun and song yibing and chen qifeng journal proceedings of the ieee cvf conference on computer vision and pattern recognition cvpr year 2025 preview_fit contain if you have any questions please contact me copyright 2026 hongyu liu powered by jekyll with al folio theme hosted by github pages photos from unsplash