LogoThread Easy
  • 探索
  • 撰写 Thread
LogoThread Easy

您的一体化 Twitter 线程助手

© 2025 Thread Easy All Rights Reserved.

探索

最新在前,按卡片方式浏览线程

开启时会模糊预览图,关闭后正常显示

Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding

• We propose Motion4D, a model that integrates 2D priors from foundation models into a dynamic 3D Gaussian Splatting representation. This achieves consistent motion and semantic modeling from monocular videos.

• We design a two-part iterative optimization framework comprising:
  - Sequential optimization: updates motion and semantic fields in consecutive stages to maintain local consistency.
  - Global optimization: jointly refines all attributes to ensure long-term coherence.

• We introduce iterative motion refinement using 3D confidence maps and adaptive resampling to enhance dynamic scene reconstruction. Semantic refinement corrects 2D semantic inconsistencies through iterative updates with SAM2.

• Our Motion4D significantly outperforms both 2D foundation models and existing 3D methods in tasks such as video object segmentation, point-based tracking, and novel view synthesis.

Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding • We propose Motion4D, a model that integrates 2D priors from foundation models into a dynamic 3D Gaussian Splatting representation. This achieves consistent motion and semantic modeling from monocular videos. • We design a two-part iterative optimization framework comprising: - Sequential optimization: updates motion and semantic fields in consecutive stages to maintain local consistency. - Global optimization: jointly refines all attributes to ensure long-term coherence. • We introduce iterative motion refinement using 3D confidence maps and adaptive resampling to enhance dynamic scene reconstruction. Semantic refinement corrects 2D semantic inconsistencies through iterative updates with SAM2. • Our Motion4D significantly outperforms both 2D foundation models and existing 3D methods in tasks such as video object segmentation, point-based tracking, and novel view synthesis.

Paper (pdf): https://t.co/POADj0bzsO Project: https://t.co/SEmM7IIDB5

avatar for MrNeRF
MrNeRF
Fri Oct 31 07:38:37
RT @arvidkahl: Tomorrow on the podcast and in the newsletter: why you should never start a SaaS business.

A couple of really painful lesso…

RT @arvidkahl: Tomorrow on the podcast and in the newsletter: why you should never start a SaaS business. A couple of really painful lesso…

Building https://t.co/od97B0HVrk and https://t.co/666FnyVVE0 in Public. Raising all the boats with kindness. 🎙️ https://t.co/6w69DZmi8H · ✍️ https://t.co/lpnor5rsTW

avatar for Arvid Kahl
Arvid Kahl
Fri Oct 31 07:38:18
Josiah now at 561 followers, orbital injection successful
I am always happy to launch another lowbie with alpha

Josiah now at 561 followers, orbital injection successful I am always happy to launch another lowbie with alpha

We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1

avatar for Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Fri Oct 31 07:37:16
RT @fi56622380: 我看到这个消息的第一感觉是,Oracle 收购SambaNova可能不太靠谱,看不出动机也没有业务上的synergy

更可能是SambaNova近况不太好,融资不顺利,估值缩水,接触Oracle算是找出路,也可以generate some bu…

RT @fi56622380: 我看到这个消息的第一感觉是,Oracle 收购SambaNova可能不太靠谱,看不出动机也没有业务上的synergy 更可能是SambaNova近况不太好,融资不顺利,估值缩水,接触Oracle算是找出路,也可以generate some bu…

Prompt Engineer, dedicated to learning and disseminating knowledge about AI, software engineering, and engineering management.

avatar for 宝玉
宝玉
Fri Oct 31 07:32:31
写技术文档时,我们常用 Markdown 格式记录,但想把这些文档发布成网站却挺麻烦,现有工具要么配置复杂,要么生成的网站加载很慢。

在 GitHub 上看到 docmd 这个开源工具,它能将我们的 Markdown 文件快速转换成美观的静态文档网站,真正做到零配置、开箱即用!

基于 Node.js 开发,只需三个简单命令就能完成从初始化到部署的全过程,生成的网站轻量且快速。

而且支持多种主题、语法高亮和深色模式,还提供提示框、卡片、步骤等组件,也可以自定义页面内容。

GitHub:https://t.co/EKAISwviVM

除此之外,还内置了 SEO、Analytics 和 Sitemap 插件,可以用来优化网站的体验。

通过 npm 全局安装后即可使用,生成的静态网站可部署到 GitHub Pages、Vercel 等任意平台。

写技术文档时,我们常用 Markdown 格式记录,但想把这些文档发布成网站却挺麻烦,现有工具要么配置复杂,要么生成的网站加载很慢。 在 GitHub 上看到 docmd 这个开源工具,它能将我们的 Markdown 文件快速转换成美观的静态文档网站,真正做到零配置、开箱即用! 基于 Node.js 开发,只需三个简单命令就能完成从初始化到部署的全过程,生成的网站轻量且快速。 而且支持多种主题、语法高亮和深色模式,还提供提示框、卡片、步骤等组件,也可以自定义页面内容。 GitHub:https://t.co/EKAISwviVM 除此之外,还内置了 SEO、Analytics 和 Sitemap 插件,可以用来优化网站的体验。 通过 npm 全局安装后即可使用,生成的静态网站可部署到 GitHub Pages、Vercel 等任意平台。

💡 挖掘开源的价值 🧑🏻‍💻 坚持分享 GitHub 上高质量、有趣、实用的教程、AI工具、前沿 AI 技术 🧐 A list cool, interesting projects of GitHub. ✏️ 公众号:GitHubDaily

avatar for GitHubDaily
GitHubDaily
Fri Oct 31 07:30:00
RT @rakyll: There is a reason why most large companies are doomed. They value complexity more than anything else. People fabricate complexi…

RT @rakyll: There is a reason why most large companies are doomed. They value complexity more than anything else. People fabricate complexi…

🇦🇺 Co-founder: @AnswerDotAI & @FastDotAI ; Prev: professor @ UQ; Stanford fellow; @kaggle president; @fastmail/@enlitic/etc founder https://t.co/16UBFTX7mo

avatar for Jeremy Howard
Jeremy Howard
Fri Oct 31 07:27:53
  • Previous
  • 1
  • More pages
  • 1613
  • 1614
  • 1615
  • More pages
  • 2127
  • Next