Meta tags:
description=LayoutGPT generates plausible image layouts and indoor scene layouts with style sheets language.;
keywords=text-to-image generation, Large Language Models, scene synthesis;

Headings (most frequently used words):

layoutgpt, scene, and, generation, for, image, 3d, synthesis, application, compositional, visual, planning, with, large, language, models, layout, based, on, text, inputs, indoor, abstract, 2d, layouts, related, links, bibtex, faithfulness, in, numerical, spatial, concepts, flexible, scenarios, completion,

Text of the page (most frequently used words):
and (21), #layoutgpt (14), visual (9), #generation (9), the (8), scene (6), image (6), text (6), language (5), models (5), llms (5), for (5), layouts (5), with (4), wang (4), performance (4), indoor (4), layout (4), numerical (4), spatial (4), inputs (4), based (3), planning (3), large (3), arxiv (3), like (3), comparable (3), synthesis (3), can (3), concepts (3), compositional (2), feng (2), weixi (2), zhu (2), wanrong (2), tsu (2), jui (2), jampani (2), varun (2), akula (2), arjun (2), xuehai (2), basu (2), sugato (2), xin (2), eric (2), william (2), yang (2), work (2), code (2), also (2), application (2), shows (2), supervised (2), methods (2), skills (2), context (2), demonstrations (2), when (2), achieves (2), users (2), multiple (2), domains (2), from (2), santa (2), webpage, built, nerfies, article, feng2023layoutgpt, title, author, journal, preprint, 2305, 15393, year, 2023, bibtex, please, check, out, previous, upon, which, build, our, framework, repository, there, lot, relevant, that, was, introduced, around, same, time, ours, glip, atiss, gligen, related, links, autoregressive, manner, enables, complete, partial, completion, conditioned, room, type, floor, plan, size, two, natural, advantages, using, attribute, binding, assign, correct, attributes, bounding, boxes, inpainting, imagine, expand, underspecified, description, certain, objects, flexible, scenarios, apply, reasoning, into, learn, through, faithfulness, combined, downstream, model, outperforms, systems, human, designing, correctness, lastly, demonstrating, its, effectiveness, potential, propose, method, compose, style, sheet, enhance, generate, plausible, ranging, images, scenes, superior, converting, challenging, relations, arrangements, faithful, attaining, high, degree, user, controllability, often, requires, intricate, fine, grained, however, such, impose, substantial, burden, compared, simple, address, issue, study, how, serve, planners, generating, conditions, thus, collaborate, generative, abstract, equal, contribution, cruz, google, barbara,

Text of the page (random words):
layoutgpt layoutgpt compositional visual planning and generation with large language models weixi feng 1 wanrong zhu 1 tsu jui fu 1 varun jampani 2 arjun akula 2 xuehai he 3 sugato basu 2 xin eric wang 3 william yang wang 1 1 uc santa barbara 2 google 2 uc santa cruz equal contribution arxiv code layoutgpt for image layout generation based on text inputs layoutgpt for 3d indoor scene synthesis abstract attaining a high degree of user controllability in visual generation often requires intricate fine grained inputs like layouts however such inputs impose a substantial burden on users when compared to simple text inputs to address the issue we study how large language models llms can serve as visual planners by generating layouts from text conditions and thus collaborate with visual generative models we propose layoutgpt a method to compose in context visual demonstrations in style sheet language to enhance the visual planning skills of llms layoutgpt can generate plausible layouts in multiple domains ranging from 2d images to 3d indoor scenes layoutgpt also shows superior performance in converting challenging language concepts like numerical and spatial relations to layout arrangements for faithful text to image generation when combined with a downstream image generation model layoutgpt outperforms text to image models systems by 20 40 and achieves comparable performance as human users in designing visual layouts for numerical and spatial correctness lastly layoutgpt achieves comparable performance to supervised methods in 3d indoor scene synthesis demonstrating its effectiveness and potential in multiple visual domains 2d image layouts faithfulness in numerical and spatial concepts layoutgpt can apply the numerical reasoning skills of llms into layout generation and learn spatial concepts through in context demonstrations flexible application scenarios two natural advantages of using llms for image layout generation 1 attribute binding assign correct attributes to the bounding boxes 2 text based inpainting imagine and expand the underspecified description of certain objects 3d scene synthesis layoutgpt shows comparable performance as supervised methods in indoor scene generation conditioned on room type and floor plan size application scene completion the autoregressive manner of llms enables layoutgpt to complete a partial scene related links please check out previous work like gligen atiss and glip upon which we build our layoutgpt framework and code repository there s also a lot of relevant work that was introduced around the same time as ours bibtex article feng2023layoutgpt title layoutgpt compositional visual planning and generation with large language models author feng weixi and zhu wanrong and fu tsu jui and jampani varun and akula arjun and he xuehai and basu sugato and wang xin eric and wang william yang journal arxiv preprint arxiv 2305 15393 year 2023 the webpage is built based on nerfies