legatoninja

One day, I will bring my gun, and find that place, where I can see the milky way.

[PrismaX @PrismaXai]

Hard lighting cut our expert-only π0.5 checkpoint's success roughly in half. The same lights cost the full fine-tune nothing.

New from @castorhat: the fine-tuning ablations behind our IROS demo. https://t.co/JciqcCBMYy

[Jim Fan @DrJimFan]

We built a robotics data engine with Astra-level performance at 125x lower cost. And we are open-sourcing the weights today! Introducing Flora, a suite of open models to parse the physical world and turn raw footage into rich training signals.

Flora already powers our in-house data engine, labeling 200K hours of videos a day on 500 GPUs. We created it to scale the frontier for robot learning. Now you own the weights too.

Flora gives you:

- High-quality, timestamped captions to curate and filter human egocentric and robot data;
- Dense progress rewards for quality control and RL post-training;
- Rubrics scoring for real-world benchmarking

Try it today, links in thread:

[The Humanoid Hub @TheHumanoidHub]

X Square Robot has introduced TwinDEX: a co-designed pair of three-finger, 9-DoF dexterous hands, one wearable for data collection, one matched robot end effector for deployment.

The wearable and the robot hand share the same kinematics, geometry, materials, look, and sensors, so human demos map straight to the robot without retargeting.

Human data collection improves throughput 5.3x compared to teleoperation, and a policy can be trained with a few hundred episodes without any robot data.

[The Humanoid Hub @TheHumanoidHub]

UMI, but for humanoid hands.

UMI-style capture already works great for simple grippers. For a high-DOF hand, where the action space is huge and contact force carries the important signal, matching the data collector to the hand matters even more.

Chestnut just announced that pair:

Aero Hand, an 18-DOF human-sized dexterous hand (tendon/linkage hybrid, tested for 3M+ cycles), and

Aero UMI, a wearable exoskeleton whose kinematics exactly matches the robotic hand, with direct per-joint angle sensing accurate to 0.1°.

Zero embodiment gap for human data collection. Doubles as a teleop rig.

[Ryohei Sasaki@engineer @rsasaki0109]

aprilcube
Generate 3D-printable cubes/cuboids with ArUco or AprilTag fiducial markers on all 6 faces, then detect their 6-DOF pose from a camera.
https://t.co/6mxt0B5ElF

[Qineng Wang @qineng_wang]

GPT-6 Astra can solve some seriously tricky spatial constraint tasks: from unlocking interlocked parts to threading a rope through three rings.

Try it yourself now 👇
https://t.co/him6NbTI2a

[Arata Horie @ArataHorie]

Awesome-UMI を公開しました！
特にヒューマノイドロボットの領域でボトルネックと言われている人間のデモンストレーションデータを収集する装置、「UMI（Universal Manipulation Interface）」系の装置をまとめたウェブサイトです。
まだ完璧ではないベータ版ですが、継続的に追加できればと思います

[legatoninja @stevenyfy]

@danfei_xu Can Insta360 provide accurate end-effector 6DoF pose estimation? Is the accuracy only at the centimeter level? Is this sufficient for pre-training?

[KK.aWSB @KKaWSB]

突发｜GitHub榜首”穿墙透视”项目火了

用家用路由器WiFi信号，就能隔墙追踪你的一举一动。

不需摄像头，不需额外硬件。

原理：分析WiFi信号在人体上的反射，AI还原24个身体部位姿态。

准确率接近摄像头。

你家的WiFi，是保护隐私的屏障，还是暴露隐私的窗口？

[Nic Cruz Patane @niccruzpatane]

Here’s what happens when a couple of Waymo vehicles meet on a narrow street:

[Danfei Xu @danfei_xu]

Aloha! I’ll be giving a talk on some of our recent works in robot learning from human data at 9am tomorrow!

[Marwa ElDiwiny @MarwaEldiwiny]

Tesla Optimus Ankle Design

Paper discussed in this episode "A Framework for Optimal Ankle Design of Humanoid Robots" https://t.co/ByjjJTlAD6

Video: https://t.co/KdG5vmCx4V

@GoingBallistic5

[Nathie @nathie]

Hand Physics Lab is coming to Apple Vision Pro. Can’t wait to jump back into this sandbox!

[Chenhao Li @breadli428]

Finally got to see the real @UnitreeRobotics R1 demo. Impressive hardware performance consistency — side flip 10 times in a row without much degradation.

[legatoninja @stevenyfy]

@ryan_punamiya You are using a very small clothes when you fold with glasses, may I ask why?

[Aran Komatsuzaki @arankomatsuzaki]

TikTok presents Depth Anything V2

Trained from 595K synthetic labeled images and 62M+ real unlabeled images, providing the most capable monocular depth estimation model

proj: https://t.co/KaOQauiOST
abs: https://t.co/9HxIpsPWJJ

[Chris Paxton @chris_j_paxton]

The longer reach and mobile base seems like it will make a difference even for laundry folding

[Tesla Optimus @Tesla_Optimus]

Poptimus

[legatoninja @stevenyfy]

@FoxNews you are doing the right thing.

[legatoninja @stevenyfy]

@TakiVan1 @TheHumanoidHub what is the usecase for this gesture?