Meta“分割一切”的SAM 3D 来了！

Ofweek维科网12-01

就在刚刚，Meta MSL 实验室发布三维重建模型 SAM 3D。“分割一切”现在可以直接生成 3D 模型了。前不久，Meta 发表 SAM 3 论文，能够检测、分割和跟踪图像和视频中的对象，还支持简短的文本短语和示例提示。现在，SAM 3 更是带来 3D 新范式，将模型引入到三维空间，从单个 2D 图像中精确重建 3D 物体和人物。从静态到3D此次发布包含两个全新模型：SAM 3D ...

网页链接

免责声明：本文观点仅代表作者个人观点，不构成本平台的投资建议，本平台不对文章信息准确性、完整性和及时性做出任何保证，亦不对因使用或信赖文章信息引发的任何损失承担责任。

精彩评论

我们需要你的真知灼见来填补这片空白

发表看法

{"i18n":{"language":"zh_CN"},"isChannel":false,"data":{"share":"https://www.laohu8.com/m/news/2588723196?lang=zh_CN&edition=full","thumbnail":"","is_english":false,"pubTime":"2025-12-01 17:07","share_image_url":"https://static.laohu8.com/9a95c1376e76363c1401fee7d3717173","id":"2588723196","market":"us","top_or_hot":-1,"title":"Meta“分割一切”的SAM 3D 来了！","media":"Ofweek维科网","content":"<div>\n<p>就在刚刚，Meta MSL 实验室发布三维重建模型 SAM 3D。“分割一切”现在可以直接生成 3D 模型了。前不久，Meta 发表 SAM 3 论文，能够检测、分割和跟踪图像和视频中的对象，还支持简短的文本短语和示例提示。现在，SAM 3 更是带来 3D 新范式，将模型引入到三维空间，从单个 2D 图像中精确重建 3D 物体和人物。从静态到3D此次发布包含两个全新模型：SAM 3D ...</p>\n\n<a href=\"http://gu.qq.com/resources/shy/news/detail-v2/index.html#/?id=nesSN20251201171116a71c89fc&s=b\">网页链接</a>\n\n</div>\n","source":"tencent","html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>Meta“分割一切”的SAM 3D 来了！</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\nMeta“分割一切”的SAM 3D 来了！\n</h2>\n\n<h4 class=\"meta\">\n\n\n2025-12-01 17:07 北京时间&nbsp;&nbsp;&nbsp;<a href=http://gu.qq.com/resources/shy/news/detail-v2/index.html#/?id=nesSN20251201171116a71c89fc&s=b><strong>Ofweek维科网</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>就在刚刚，Meta MSL 实验室发布三维重建模型 SAM 3D。“分割一切”现在可以直接生成 3D 模型了。前不久，Meta 发表 SAM 3 论文，能够检测、分割和跟踪图像和视频中的对象，还支持简短的文本短语和示例提示。现在，SAM 3 更是带来 3D 新范式，将模型引入到三维空间，从单个 2D 图像中精确重建 3D 物体和人物。从静态到3D此次发布包含两个全新模型：SAM 3D ...</p>\n\n<a href=\"http://gu.qq.com/resources/shy/news/detail-v2/index.html#/?id=nesSN20251201171116a71c89fc&s=b\">网页链接</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","isBrief":false,"type":0,"news_type":1,"symbol":"LU1489326972.SGD","symbol_name":"First Eagle Amundi International AHS-MD SGD-H","start_time":0,"source_url":"http://gu.qq.com/resources/shy/news/detail-v2/index.html#/?id=nesSN20251201171116a71c89fc&s=b","article_id":"2588723196","we_media_id":null,"thumbnails":[],"rights":{"source":"tencent","url":"http://gu.qq.com/resources/shy/news/detail-v2/index.html#/?id=nesSN20251201171116a71c89fc&s=b","rn_cache_url":null,"customStyle":"body{padding-top:10px;}#news_title{font-weight:bold;#titleStyle#;}#news_description span{font-size:12px;#descriptionStyle#;}.footer-note{#statement#}","selectors":".mod-LoadTzbdNews, body","filters":".relate-stock, .hot-list, .recom-box, .wx-sou","directOrigin":true},"url":"https://stock-news.laohu8.com/highlight/detail?id=2588723196","pubTimestamp":1764580020,"columns":[],"sourceInfo":{"source_id":"tencent","name":"腾讯"},"weMediaInfo":null,"summary":"就在刚刚，Meta MSL 实验室发布三维重建模型 SAM 3D。“分割一切”现在可以直接生成 3D 模型了。前不久，Meta 发表 SAM 3 论文，能够检测、分割和跟踪图像和视频中的对象，还支持简短的文本短语和示例提示。SAM 3D Objects 利用强大的数据标注引擎，将 3D 数据和训练方案紧密结合。SAM 3D Objects 能够很好地进行泛化，并支持密集场景重建。SAM 3D Body 利用大规模、高质量数据和稳健的训练策略，提供精准且鲁棒的 3D 人体姿态和形状估计。","collect":0,"end_time":0,"defaultTopTitle":"qq.com","property":[],"viewcount":null,"language":"zh","relate_stocks":{"LU1489326972.SGD":"First Eagle Amundi International AHS-MD SGD-H","LU1732800096.USD":"摩根大通环球收益基金A (irc)","LU0077335932.USD":"FIDELITY AMERICAN GROWTH \"A\" INC","FB":"ProShares S&P 500 Dynamic Buffer ETF","LU0006306889.USD":"SCHRODER ISF US LARGE CAP \"A\" (USD) INC AV","BK4550":"红杉资本持仓","LU0719512351.SGD":"JPMorgan Funds - US Technology A (acc) SGD","LU1066051498.USD":"HSBC GIF GLOBAL EQUITY VOLATILITY FOCUSED \"AM2\" (USD) INC","LU1571399168.USD":"ALLSPRING GLOBAL LONG/SHORT EQUITY \"IP\" (USD) ACC","LU1035775433.USD":"AB SICAV I - AMERICAN GROWTH PORTFOLIO \"AD\" (USD) INC","LU0203201768.USD":"AB SICAV I - ALL MARKET INCOME PORTFOLIO \"AX\" (USD) INC","BK4507":"流媒体概念","LU0683600562.USD":"AB SELECT US EQUITY \"A\" (USD) ACC","LU1084165304.USD":"FIDELITY WORLD \"A\" (USD) ACC","BK4551":"寇图资本持仓","LU0444971666.USD":"天利全球科技基金","META":"Meta Platforms, Inc.","LU0265550946.USD":"BGF SYSTEMATIC GLOBAL ENHANCED EQUITY YIELD \"A5\" (USD) INC","SAM":"波斯顿啤酒","LU2552382058.USD":"WELLINGTON US BRAND POWER \"A\" (USD) ACC","LU1127390331.HKD":"AB SICAV I - ALL MARKET INCOME PORTFOLIO \"A\" (HKD) ACC","LU1551013342.USD":"Allianz Income and Growth Cl AMg2 DIS USD","LU0354030438.USD":"富国美国大盘成长基金Cl A Acc","IE00BK4W5L77.USD":"HSBC GLOBAL FUNDS ICAV US EQUITY INDEX \"HC\" (USD) ACC","LU2552382132.HKD":"WELLINGTON US BRAND POWER \"A\" (HKD) ACC","LU1868836757.USD":"CT (LUX) I AMERICAN FUND \"2\" (USD) ACC","BK4585":"ETF&股票定投概念","IE00B1XK9C88.USD":"PINEBRIDGE US LARGE CAP RESEARCH ENHANCED \"A\" (USD) ACC","IE00BQXX3D17.EUR":"GUINNESS GLOBAL INNOVATORS \"C\" (EUR) ACC","LU2265009873.SGD":"Eastspring Investments - Global Growth Equity AS SGD-H","LU2404859741.USD":"FUNDSMITH EQUITY FUND \"R\" (USD) INC","LU2087621335.USD":"ALLSPRING GLOBAL FACTOR ENHANCED EQUITY \"A\" (USD) ACC","SGXZ51526630.SGD":"大华环球创新基金A Acc SGD","LU2458330169.SGD":"FRANKLIN SHARIAH TECHNOLOGY \"A\" (SGD) ACC","LU2247934214.USD":"FIDELITY FUNDS SUSTAINABLE FUTURE CONNECTIVITY \"A\" (USD) ACC","IE0003U64NQ7.SGD":"PIMCO BALANCED INCOME AND GROWTH \"M\" (SGDHDG) ACC","LU2404859667.USD":"FUNDSMITH EQUITY FUND \"R\" (USD) ACC","LU0648001328.SGD":"Natixis Harris Associates US Equity RA SGD","LU0276348264.USD":"THREADNEEDLE (LUX) GLOBAL DYNAMIC REAL RETURN\"AUP\" (USD) INC","BK4587":"ChatGPT概念","IE000KEQY171.SGD":"PIMCO BALANCED INCOME AND GROWTH \"M\" (SGDHDG) INC","LU0820562030.AUD":"ALLIANZ INCOME AND GROWTH \"AMH2\" (AUDHDG) H2 INC","LU2458330243.SGD":"FRANKLIN SHARIAH TECHNOLOGY \"A-H1\" (SGDHDG) ACC","LU1069344957.HKD":"AB SICAV I - AMERICAN GROWTH PORTFOLIO \"AD\" (HKD) INC","LU0345769631.USD":"NINETY ONE GSF GLOBAL EQUITY \"A\" (USD) INC","LU0690374615.EUR":"FUNDSMITH EQUITY \"R\" (EUR) ACC","SGXZ31699556.SGD":"UGDP UNITED GLOBAL QUALITY GROWTH \"C\" (SGDHDG) ACC","LU0690374961.EUR":"FUNDSMITH EQUITY \"R\" (EUR) INC","IE0009356076.USD":"JANUS HENDERSON GLOBAL TECHNOLOGY AND INNOVATION \"A2\" (USD) ACC","LU0203202063.USD":"AB SICAV I - ALL MARKET INCOME PORTFOLIO \"A2X\" (USD) ACC"},"translate_title":"Meta's \"Split Everything\" SAM 3D is here!","themeId":null,"isJumpTheme":false,"ttsUrl":null,"symbols_score_info":{"FB":1.5,"META":0.9,"SAM":1.5},"content_text":"就在刚刚，Meta MSL 实验室发布三维重建模型 SAM 3D。“分割一切”现在可以直接生成 3D 模型了。前不久，Meta 发表 SAM 3 论文，能够检测、分割和跟踪图像和视频中的对象，还支持简短的文本短语和示例提示。现在，SAM 3 更是带来 3D 新范式，将模型引入到三维空间，从单个 2D 图像中精确重建 3D 物体和人物。从静态到3D此次发布包含两个全新模型：SAM 3D Objects 模型：用于物体和场景重建SAM 3D Body 模型：用于人体和体型估计两个模型均能够将静态 2D 图像转换为精细的 3D 重建模型。3D ObjectsSAM 3D Objects 能够从单张自然图像中实现稳定且基于视觉的 3D 重建和物体姿态估计，从而从日常图像中重建物体的精细 3D 形状、纹理和布局。针对小物体、间接视角和遮挡现象，模型可以通过识别上下文信息辅助重建，弥补像素不足的缺点。使用 SAM 3D Objects，用户可以从一张图像开始选择任意物体，快速生成 3D 模型。这意味着用户可以在重建的 3D 场景中精确操作单个物体，从不同视角进行观察。过去的 3D 模型受到数据可用性的限制，真实数据的数量级有限，导致 3D 模型的应用范围仅限于合成或人为设置的场景。要应用于日常环境中常见的更具挑战性的场景，则需要一种新的方法。SAM 3D Objects 利用强大的数据标注引擎，将 3D 数据和训练方案紧密结合。与文本、图像或视频不同，从零开始创建 3D 真实数据需要高度专业化的技能，3D 数据采集效率低、成本高。然而，验证或对网格进行排序更简单。可以通过构建一个数据引擎来实现规模化。该引擎要求标注员对循环模型套件生成的多个选项进行评分，最难的部分可交由 3D 美术师完成，以填补数据盲区。这样，研究人员就能大规模对物理世界图像进行标注，包括 3D 对象的形状、纹理和布局。SAM 3D 一共标注了近 100 万张不同的图像，并生成了约 314 万个循环模型网格。为了使模型能够处理自然图像，后续的后训练阶段需要进行校准。数据引擎还能提供数据来支持后训练过程。模型鲁棒性和输出质量的提升也使数据引擎能够更好地生成数据，从而形成一个正反馈循环。为搭建单张图像三维重建物理世界物体的自然图像分布基准数据集，Meta 还构建了SAM 3D 艺术家物体数据集（SA-3DAO）。这是首个用于评估基于视觉的物理世界图像三维重建的数据集，包含更具挑战性的多样化图像和物体。SAM 3D Objects 能够很好地进行泛化，并支持密集场景重建。在与人类用户进行的直接对比测试中，其成功率至少达到其他领先模型的 5 倍。通过扩散捷径和工程优化，可以在几秒钟内生成高质量重建结果。官方表示，下一步是提高输出分辨率和改进物体布局。目前，输出分辨率仅限“适中”水平，限制了复杂物体的细节呈现，可能会出现失真或丢失细节。因此，下一步自然是提高输出分辨率。SAM 3D Objects 目前一次只能预测一个物体，并没有接受过物理交互的推理训练。如果能够预测多个物体并结合适当的损失函数，就可以对场景中的多个物体进行联合推理。3D BodySAM 3D Body 即使在涉及不寻常姿态、图像部分被遮挡或多人等复杂情况下，也能精确估计 3D 人体姿态和形状的需求。该模型利用了一种名为Meta Momentum Human Rig (MHR)的新型开源 3D 网格格式，通过分离人体骨骼结构和软组织形状提高可解释性。基于 Transformer 编码器-解码器架构，模型可以预测 MHR 网格参数——图像编码器采用多输入设计来捕捉人体部位的高分辨率细节，而网格解码器则扩展以支持基于提示的预测。SAM 3D Body 利用大规模、高质量数据和稳健的训练策略，提供精准且鲁棒的 3D 人体姿态和形状估计。研究人员构建了一个包含数十亿张图像的大型数据集，并使用可扩展的自动化数据引擎挖掘高价值图像。他们构建了一个包含约800 万张图像的高质量训练数据集，用于训练模型，使其能够应对遮挡、罕见姿态和各种服装。下一步，研究团队表示会将人、物体和环境之间的交互纳入模型训练。还会继续提升手部姿态估计的性能，目前其准确率仍不及专门的手部姿态估计方法。参考资料：https://ai.meta.com/blog/sam-3d/?utm_source=twitter&utm_medium=organic_social&utm_content=video&utm_campaign=sam","kind":"news","is_publish_news":true,"is_publish_highlight":false,"is_publish_live":false,"is_publish_wemedia":null,"editions":null,"column":"","sentiment":"0","news_tag":"","news_rank":0,"symbols":[],"gpt_button":1,"need_auth":false,"code":"91000000","status":"200"},"commentList":[],"isCommentEnd":true,"newsSizeData":{"likeSize":0,"commentSize":0,"repostSize":0,"favoriteSize":0,"likeStatus":false,"favoriteStatus":false},"APP":{"userAgent":"Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)","isDev":false,"isTTM":false,"deviceId":"web-server-community-laohu8-v3","version":"4.35.9","shortVersion":"4.35.9","platform":"web","vendor":"web","appName":"laohu8","isIOS":false,"isAndroid":false,"isTiger":false,"isTHS":false,"isWeiXin":false,"isWeiXinMini":false,"isWeiBo":false,"isQQ":false,"isBaiduSwan":false,"isBaiduBox":false,"isDingTalk":false,"isToutiao":false,"isOnePlus":false,"isHuaWei":false,"isXiaomi":false,"isXiaomiWebView":false,"isOppo":false,"isVivo":false,"isSamsung":false,"isMobile":false},"href":"/m/news/2588723196","isCrawlerRequest":true}