Discover

Explore inspiring content shared by robotics researchers, engineers, and enthusiasts from around the world.

27 Articles
2 Experiments
0 Datasets
0 Failures
🎯
Initiate #1 Meetup
Upcoming
📅 线下聚会

Initiate #1 · XRollout 首届社区聚会

X XRollout Team

众人拾柴火焰高。开源机器人智能社区首次线下聚会,欢迎机器人研究者、工程师、硬件爱好者一起交流,共同探讨开放机器人生态的未来。

Date April 2026
Location Shenzhen
🤖
Agent AI Program
Free Tokens
🎓 Program

Agent AI Skill Developer Program

X XRollout AI Team

Learn to develop custom skills based on Agent AI frameworks, from fundamental concepts to complete projects. Earn free token credits for robot model training after completion.

Agent AI Claude Code Free Credits
👥
Human Data Collection
📖 Methodology

Human Data Collection Pipeline

X XRollout Team

A cost-effective, distributed approach to scaling up human demonstration data for robot learning. Collect thousands of trajectories at 100× lower cost than traditional in-lab collection.

Imitation Learning Data Collection Crowdsourcing
🎬
ScreenCast Pro
100% Free
🛠️ Tool

ScreenCast Pro — 在线录屏工具

X XRollout Team

Professional browser-based screen recording tool. Record experiments, demos, and tutorials directly in browser - no installation required. Completely free and privacy-focused.

Screen Recorder Utility Free
📄
Article

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

D by david

DeepSeek-V4系列预览版论文中文翻译:1.6T参数MoE模型,支持百万Token上下文,引入CSA/HCA混合注意力、mHC超连接和Muon优化器。

📄
Article

计算机视觉的惨痛教训(The Bitter Lesson of Computer Vision)

D by david

Vincent Sitzmann 关于计算机视觉未来的深度思考:传统的中间表示(如三维重建、分割掩码)将变得过时,计算机视觉的未来是作为端到端感知-动作循环的一部分。

📄
Article

《VGGT》与《SwiftVGGT》深度解读:视觉几何基础模型的统一多任务范式

D by david

深度解读 CVPR 2025 Best Paper VGGT 及其后续工作 SwiftVGGT。VGGT 通过单次前向传播同时输出相机参数、深度图、点云和点跟踪,精度超越传统优化方法。SwiftVGGT 在此基础上通过单步 SVD 和内置回环检测,将大规模场景重建速度提升 3 倍,且无需任何训练。

📄
Article

具身智能公司:本体-小脑路线 vs 大脑优先路线

D by david

A deep dive into the two major strategic paths in the embodied AI industry: Cerebellum-First vs. Brain-First, and the companies leading each camp.

📄
Article

《LingBot-Map: 用于实时三维重建的几何上下文 Transformer》深度解读

D by david

LingBot-Map 是一种基于 Transformer 的前馈式三维基础模型,实现了超长序列的高精度、实时单目三维重建与位姿估计。

📄
Article

π0.7:具有涌现能力的可引导通用机器人基础模型(中文翻译)

D by david

Physical Intelligence 最新论文 π0.7 中文全文翻译 - 一个5B参数的可引导通用机器人基础模型,展现出组合泛化能力,能够开箱即用地执行复杂灵巧任务,实现零样本跨具身迁移。

📄
Article

π0.7:具有涌现能力的可引导通用机器人基础模型(中文全译)

D by david

本文是Physical Intelligence最新论文π0.7的完整中文翻译。π0.7是一个可引导的通用机器人基础模型,能够在开箱即用情况下执行高度灵巧的长周期任务,实现零样本跨具身迁移,并能通过语言指导学习新任务。

📄
Article

《SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control》深度解读

D by david

NVIDIA 研究员将缩放定律(Scaling Law)应用到人形机器人全身控制,通过将模型从 1.2M 放大到 42M 参数、数据集扩大到 1 亿帧(700 小时),得到一个通用人形机器人基础控制器,支持多种输入接口(VR遥操作、视频、VLA模型)。

Experiment

ORB-SLAM (纯Python实现) - TUM fr1_xyz 数据集测试

D by david

ORB-SLAM Baseline Experiment This experiment tests a traditional ORB-based SLAM approach using pure Python + OpenCV...

Monocular Camera
Experiment

Camera Calibration Pipeline Test - Synthetic Chessboard

D by david

Testing the new camera calibration pipeline with automatic quality evaluation. This test uses a synthetically...

Smart Roller