{"id":9994,"date":"2026-07-08T17:04:15","date_gmt":"2026-07-08T09:04:15","guid":{"rendered":"https:\/\/infernews.com\/blog\/unifying-3d-scene-generation-and-reconstruction-in-pixel-space\/"},"modified":"2026-07-08T17:05:53","modified_gmt":"2026-07-08T09:05:53","slug":"unifying-3d-scene-generation-and-reconstruction-in-pixel-space","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/unifying-3d-scene-generation-and-reconstruction-in-pixel-space\/","title":{"rendered":"PixWorld \u628a 3D \u91cd\u5efa\u8207\u751f\u6210\u6536\u5728\u540c\u4e00\u8def\u7dda"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/Speed.jpg\" alt=\"PixWorld teaser\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u505a 3D \u5834\u666f\u6642\uff0c\u5f88\u591a\u65b9\u6cd5\u6703\u628a\u91cd\u5efa\u540c\u751f\u6210\u5206\u958b\u8655\u7406\uff0cPixWorld \u5c31\u53cd\u5176\u9053\u800c\u884c\uff0c\u628a\u5169\u8005\u6536\u9032\u540c\u4e00\u500b\u958b\u6e90\u6846\u67b6\u5f0f\u7814\u7a76\u9805\u76ee\uff1a\u540c\u4e00\u500b end-to-end pixel-space diffusion model\uff0c\u540c\u4e00\u8d9f forward pass\uff0c\u540c\u6642\u517c\u9867 reconstruction \u540c generation\u3002\u5c0d\u60f3\u6bd4\u8f03 text\u21923D\u3001image\u21923D \u540c\u591a\u8996\u89d2\u91cd\u5efa\u6d41\u7a0b\u7684\u4eba\u4f86\u8aaa\uff0c\u9019\u7a2e\u8a2d\u8a08\u6700\u5927\u610f\u7fa9\u5728\u65bc\u5c11\u4e00\u5c64\u8f49\u63db\uff0c\u4e5f\u5c11\u4e00\u5957\u5f7c\u6b64\u5272\u88c2\u7684\u6a21\u578b\u5fc3\u667a\u8ca0\u64d4\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u6279\u8a55\u7684\u820a\u7bc4\u5f0f\u5f88\u660e\u78ba\uff1a\u4e0d\u5c11\u505a\u6cd5\u5148\u9032 latent space\uff0c\u8981\u7d93\u904e VAE\/RAE\uff0c\u518d\u7531 3D decoder \u9084\u539f\u6210 3D \u8868\u793a\uff1bPixWorld \u5247\u76f4\u63a5\u5728 rendered multi-view images \u4e0a\u65bd\u52a0 pixel-space flow-matching loss\uff0c\u4ee4\u512a\u5316\u76ee\u6a19\u66f4\u8cbc\u8fd1 3D scene fidelity\uff0c\u800c\u5514\u4fc2\u5148\u5c0d\u6e96\u4e2d\u4ecb latent target\u3002\u4f5c\u8005\u518d\u52a0\u4e0a\u4e00\u500b geometry perception loss\uff0c\u501f\u52a9 frozen 3D foundation model \u7684 \u03c0\u00b3 \/ VGGT feature space\uff0c\u88dc\u56de\u7d14 2D photometric loss \u540c perceptual loss \u5c0d 3D \u7d50\u69cb\u76e3\u7763\u4e0d\u8db3\u7684\u554f\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u6703\u628a posed multi-view inputs \u5206\u6210 clean views \u540c noisy views\uff1a\u524d\u8005\u8d70 reconstruction\uff0c\u5f8c\u8005\u8d70 generation\uff0c\u4e26\u7531 two-stream diffusion transformer \u8655\u7406\uff0c\u6700\u5f8c\u89e3\u78bc\u6210 pixel-aligned 3D Gaussian representation\u3002\u9019\u500b\u53d6\u5411\u7684\u597d\u8655\uff0c\u662f\u628a 3D reconstruction\u3001text\u21923D \u540c image\u21923D \u653e\u5230\u540c\u4e00\u500b\u8868\u793a\u8207\u8a13\u7df4\u6846\u67b6\uff1b\u4ee3\u50f9\u5247\u662f\u9805\u76ee\u76ee\u524d\u4ecd\u504f\u7814\u7a76\u539f\u578b\uff0cREADME \u5df2\u63d0\u5230 cleaned RealEstate10K\u3001DL3DV\u3001ACID datasets \u4ee5\u53ca PixWorld-480P-4steps distilled weights \u8207 inference code \u5c1a\u672a\u5b8c\u6574\u91cb\u51fa\u3002<\/p>\n\n\n<div\n    \tclass=\"align wp-block-vpb-video\"    id='vpbpVideoPlayer-1'\n    data-attributes='{&quot;source&quot;:&quot;https:\\\/\\\/sensengao.github.io\\\/PixWorld\\\/static\\\/videos\\\/demo_v32.mp4&quot;,&quot;repeat&quot;:true,&quot;autoplay&quot;:true,&quot;muted&quot;:true,&quot;align&quot;:&quot;&quot;,&quot;poster&quot;:&quot;&quot;,&quot;controls&quot;:{&quot;play-large&quot;:true,&quot;restart&quot;:false,&quot;rewind&quot;:true,&quot;play&quot;:true,&quot;fast-forward&quot;:true,&quot;progress&quot;:true,&quot;current-time&quot;:true,&quot;duration&quot;:false,&quot;mute&quot;:true,&quot;volume&quot;:true,&quot;pip&quot;:false,&quot;airplay&quot;:false,&quot;captions&quot;:false,&quot;settings&quot;:true,&quot;download&quot;:false,&quot;fullscreen&quot;:true},&quot;captions&quot;:[],&quot;ratio&quot;:&quot;16:9&quot;,&quot;accentColor&quot;:&quot;&quot;,&quot;lazyLoad&quot;:false,&quot;width&quot;:&quot;100%&quot;,&quot;radius&quot;:&quot;0px&quot;,&quot;resetOnEnd&quot;:false,&quot;autoHideControl&quot;:true,&quot;isSetup&quot;:false}'\n><\/div>\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u985e\u578b\u5b9a\u4f4d<\/strong>\uff1a\u5c6c\u65bc\u6a21\u578b\u5c0e\u5411\u7684\u958b\u6e90\u9805\u76ee\uff0c\u8655\u7406 3D scene generation \u8207 reconstruction \u5206\u88c2\u6210\u5169\u5957\u6d41\u7a0b\u7684\u554f\u984c\u3002<\/li>\n\n\n\n<li><strong>\u4e3b\u8981\u5dee\u7570<\/strong>\uff1a\u76f4\u63a5\u5728 pixel space \u8a13\u7df4\uff0c\u907f\u958b VAE\/RAE \u4e2d\u4ecb\u8868\u793a\uff0c\u4e26\u52a0\u5165 geometry-aware \u76e3\u7763\u3002<\/li>\n\n\n\n<li><strong>\u5df2\u77e5\u8868\u73fe<\/strong>\uff1adistilled 4-step \u7248\u672c\u6bcf\u500b scene \u7d04 0.6 \u79d2\uff0c\u5b98\u65b9\u7a31\u76f8\u5c0d diffusion-based world generators \u5cf0\u503c\u53ef\u9054\u7d04 1000\u00d7 \u52a0\u901f\u3002<\/li>\n\n\n\n<li><strong>\u91cd\u5efa\u6307\u6a19<\/strong>\uff1a\u63d0\u4f9b 71.04 WorldScore average\uff0c\u4ee5\u53ca RealEstate10K 4-view reconstruction \u7684 26.21 dB PSNR\u3002<\/li>\n\n\n\n<li><strong>\u76f8\u95dc\u6a21\u578b\u8207\u6a21\u7d44<\/strong>\uff1atwo-stream diffusion transformer\u3001pixel-aligned 3D Gaussian\u3001frozen 3D foundation model \u03c0\u00b3 \/ VGGT\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u968e\u6bb5\u8f03\u9069\u5408\u7814\u7a76 3D world generation\u30013D Gaussian Splatting\u3001multi-view reconstruction \u7684\u5718\u968a\u5148\u7528\u4f86\u5224\u65b7\u65b9\u6cd5\u50f9\u503c\uff0c\u800c\u5514\u4fc2\u5373\u523b\u7576\u6210\u5b8c\u6574\u751f\u7522\u5de5\u5177\u90e8\u7f72\u3002\u53ef\u60dc\u76ee\u524d\u516c\u958b\u8cc7\u8a0a\u4ecd\u4ee5\u8ad6\u6587\u3001\u793a\u7bc4\u9801\u8207\u65b9\u6cd5\u4ecb\u7d39\u70ba\u4e3b\uff0c\u672a\u898b\u5b8c\u6574\u5b89\u88dd\u3001\u8a55\u6e2c\u8173\u672c\u8207\u6b0a\u91cd\u767c\u5e03\uff0c\u56e0\u6b64\u8f03\u5408\u7406\u7684\u7406\u89e3\u65b9\u5f0f\uff0c\u662f\u628a PixWorld \u8996\u70ba\u4e00\u689d\u5f88\u6709\u65b9\u5411\u611f\u7684\u65b0\u8def\u7dda\uff1a\u5b83\u4e0d\u53ea\u662f\u8ffd\u6c42\u66f4\u5feb\uff0c\u9084\u8a66\u5716\u91cd\u65b0\u5b9a\u7fa9 3D \u751f\u6210\u8207\u91cd\u5efa\u61c9\u8a72\u5171\u7528\u540c\u4e00\u5957\u8a13\u7df4\u76ee\u6a19\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/sensengao.github.io\/PixWorld\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/SensenGao\/PixWorld\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/arxiv.org\/pdf\/2607.05373\" rel=\"noopener noreferrer\"><strong>Paper<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u60f3\u7531\u5e7e\u5f35\u8996\u89d2\u5716\u91cd\u5efa\u5834\u666f\uff0c\u6216\u8005\u76f4\u63a5\u505a text\u21923D\u3001image\u21923D\uff0cPixWorld \u628a\u5169\u689d\u8def\u5408\u4f75\u6210\u540c\u4e00\u500b\u6a21\u578b\u3002\u5b83\u7684\u5438\u5f15\u529b\u4e0d\u53ea\u5728\u901f\u5ea6\uff0c\u4ea6\u5728\u65bc\u8a13\u7df4\u76ee\u6a19\u66f4\u8cbc\u8fd1 3D \u5834\u666f\u672c\u8eab\u3002<\/p>\n","protected":false},"author":8,"featured_media":9993,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,30,132,76,127],"tags":[],"class_list":["post-9994","post","type-post","status-publish","format-standard","hentry","category-133","category-image","category-3d","category-76","category-127"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9994","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=9994"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9994\/revisions"}],"predecessor-version":[{"id":9996,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9994\/revisions\/9996"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9993"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9994"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9994"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9994"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}