Meta tags:
description=DESCRIPTION META TAG;
keywords=KEYWORDS SHOULD BE PLACED HERE;

Headings (most frequently used words):

objects, of, interactable, part, results, to, complex, in, cabinet, artgs, replicas, two, multi, paris, rendering, synthetic, real, world, meshes, drag, mouse, rotate, scroll, wheel, zoom, out, table, storage, comparisons, with, building, articulated, via, gaussian, splatting, top, bottom, reconstruct, articulation, from, interaction, abstract, method, overview, simulation, isaacsim, 31249, 47648, 45503, oven, 101908, 25493, whole, scene, 1r, 1r2p, microwave, 3r, washing, machine, dta, bibtex, images,

Text of the page (most frequently used words):
and (17), part (15), the (14), #objects (14), nbsp (10), complex (6), #articulated (6), artgs (6), interactable (5), multi (5), with (5), gaussians (5), for (5), this (4), page (4), results (4), our (4), articulation (4), reconstruction (4), that (3), building (3), replicas (3), two (3), cabinet (3), real (3), world (3), synthetic (3), method (3), canonical (3), text (3), single (3), module (3), mesh (3), was (2), from (2), project (2), you (2), template (2), website (2), international (2), via (2), gaussian (2), splatting (2), liu (2), jia (2), baoxiong (2), ruijie (2), junfeng (2), zhu (2), song (2), chun (2), huang (2), siyuan (2), learning (2), 2025 (2), comparisons (2), paris (2), drag (2), mouse (2), rotate (2), scroll (2), wheel (2), zoom (2), out (2), meshes (2), table (2), storage (2), rendering (2), overview (2), coarse (2), each (2), state (2), assignment (2), including (2), key (2), information (2), across (2), different (2), object (2), states (2), dynamics (2), modeling (2), approach (2), both (2), university (2), built, using, which, adopted, are, free, borrow, just, ask, link, back, footer, licensed, under, creative, commons, attribution, sharealike, license, nerfies, academic, inproceedings, liu2025building, title, author, booktitle, thirteenth, conference, representations, year, bibtex, dta, top, bottom, washing, machine, microwave, 1r2p, whole, scene, images, 25493, oven, 101908, 45503, 47648, 31249, simulation, isaacsim, divided, into, stages, obtaining, init, matching, trained, individually, initializing, clustered, centers, jointly, optimizing, model, parameters, phi, challenge, computer, vision, existing, methods, often, fail, effectively, integrate, limiting, accuracy, particularly, introduce, novel, leverages, flexible, efficient, representation, address, these, issues, incorporates, fine, initialization, updates, aligning, employs, skinning, inspired, improve, extensive, experiments, datasets, new, benchmark, demonstrate, achieves, art, performance, joint, parameter, estimation, significantly, improves, quality, efficiency, especially, additionally, provide, comprehensive, analyses, design, choices, validating, effectiveness, component, highlight, potential, areas, future, improvement, abstract, reconstruct, interaction, data, code, now, available, arxiv, paper, iclr, tsinghua, national, lab, general, bigai, peking, indicates, equal, contribution,

Text of the page (random words):
artgs artgs building interactable replicas of complex articulated objects via gaussian splatting yu liu 1 2 baoxiong jia 2 ruijie lu 2 3 junfeng ni 1 2 song chun zhu 1 2 3 siyuan huang 2 indicates equal contribution 1 tsinghua university nbsp nbsp 2 national key lab of general ai bigai nbsp nbsp 3 peking university iclr 2025 paper nbsp nbsp arxiv nbsp nbsp code now available nbsp nbsp data reconstruct articulation from interaction abstract building articulated objects is a key challenge in computer vision existing methods often fail to effectively integrate information across different object states limiting the accuracy of part mesh reconstruction and part dynamics modeling particularly for complex multi part articulated objects we introduce artgs a novel approach that leverages 3d gaussians as a flexible and efficient representation to address these issues our method incorporates canonical gaussians with coarse to fine initialization and updates for aligning articulated part information across different object states and employs a skinning inspired part dynamics modeling module to improve both part mesh reconstruction and articulation learning extensive experiments on both synthetic and real world datasets including a new benchmark for complex multi part objects demonstrate that artgs achieves state of the art performance in joint parameter estimation and part mesh reconstruction our approach significantly improves reconstruction quality and efficiency especially for multi part articulated objects additionally we provide comprehensive analyses of our design choices validating the effectiveness of each component to highlight potential areas for future improvement method overview the overview of artgs our method is divided into two stages i obtaining coarse canonical gaussians g c_ text init by matching the gaussians g 0_ text single and g 1_ text single trained with each single state individually and initializing the part assignment module with clustered centers ii jointly optimizing canonical gaussians g c and the articulation model including the articulation parameters phi and the part assignment module rendering results of synthetic objects two part objects multi part objects rendering results of real world objects simulation in isaacsim interactable meshes of complex synthetic objects drag mouse to rotate scroll wheel to zoom in out table 31249 storage 47648 storage 45503 oven 101908 table 25493 interactable meshes of complex real world objects drag mouse to rotate scroll wheel to zoom in out images interactable replicas whole scene cabinet 1r cabinet 1r2p microwave cabinet 3r washing machine comparisons with paris top results of artgs bottom results of paris comparisons with dta two part objects multi part objects bibtex inproceedings liu2025building title building interactable replicas of complex articulated objects via gaussian splatting author liu yu and jia baoxiong and lu ruijie and ni junfeng and zhu song chun and huang siyuan booktitle the thirteenth international conference on learning representations year 2025 this page was built using the academic project page template which was adopted from the nerfies project page you are free to borrow the template of this website we just ask that you link back to this page in the footer this website is licensed under a creative commons attribution sharealike 4 0 international license