Yu Xiang

@YuXiang_IRVL • 6,425 subscribers

I do research in robotics and computer vision @IRVLUTD, Ex- research scientist @NVIDIA, postdoc @uwcse and @StanfordCS, phd @UMich, all views my own.

Shorts

50,151 views

32,598 views

30,015 views

53,676 views

18,972 views

21,236 views

Videos

sweetdream.ai

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Private Show

Join now for exclusive access

Free preview available • Premium content

193,979 views • 8 months ago

109,171 views • 9 months ago

93,108 views • 8 months ago

80,153 views • 7 months ago

70,698 views • 8 months ago

29,338 views • 4 months ago

41,223 views • 7 months ago

29,826 views • 5 months ago

19,066 views • 3 months ago

57,431 views • 1 year ago

32,917 views • 11 months ago

40,329 views • 1 year ago

41,677 views • 1 year ago

18,502 views • 9 months ago

11,114 views • 5 months ago

No more content to load

Live Cam

Yu Xiang

Shorts

Temporal and spatial alignment between glove and camera data. It is not an easy task, but it’s coming together. Human data for robots. #Robotics

Left: raw depth from the OAK-D Pro W camera Right: depth from FoundationStereo

If you are interested in ego-centric data capture, no need to get a headset or smart glasses. Get a $10 head strap mount and put one of these cameras on it. You get RGB, stereo, depth, 3D point clouds, and camera trajectory. Visualization powered by Rerun. Pretty cool!

Occlusion is still a challenge for SAM2. Video object segmentation is not solved yet

Excited to share that we contributed our SceneReplica work into ManipulationNet as a task for Grasping in Clutter! ManipulationNet is a community effort for benchmarking in robotics. Evaluate your grasping algorithms or policies here:

🤖Why is robot manipulation still an open challenge? This video shows a kitting task -- packing multiple items into a single product. No robot today can do this autonomously. Big challenges = big opportunities for research and industry. #robotics #manipulation #automation

Videos

Watch Anya Live

Successfully running Ultralytics YOLOv8 on a Jetson Nano with an IMX219 camera. Real-time vision on the edge.

Our students are presenting their robotics projects this week! I’ll be sharing some of the coolest ones over the next few days. The first one is a dancing SO-101😄Turn on your volume, there’s music! #Robotics

If you’re considering a stereo camera for Jetson, I tested a ZED Mini Stereo camera. It’s about $400 (vs. $250 for Jetson Orin Nano), and Stereolabs says current lead time is 6 weeks. I ended up ordering from a third-party seller.

Handed out 21 NVIDIA Jetson Orin Nano kits to students today. I did not realize downloading the JetPack 6.2 image would take so long. Most of the time was spent just downloading or copying the image file😅Only a few teams managed to get the system set up today.

I am really impressed by how easy Spectacular AI makes visual-inertial odometry with an OAK camera. Just "pip install spectacularAI" and run a Python script. Running VIO/SLAM used to be much harder in my opinion.

We (Luis Felipe Casas and I) handed out 21 NVIDIA #Jetson Orin Nano kits to students in my computer vision class this semester. All teams showcased their live Jetson-powered demos in the classroom at the end of the course! Here are a few sample projects👇 1. Hand gesture for control

From egocentric glove data to a MANO hand model, replayed in Rerun.

We noticed that DINOv3 was surprisingly strong at matching object features. This inspired L2G (Local Matches to Global Masks). With a few reference images, a robot can search a room for the target object. 🔗 Project: 💻 Code:

Among the policies we evaluated so far (ACT, DiT, SmolVLA, pi0, pi0.5), fine-tuned pi0.5 achieves the best performance on VLA-REPLICA. The trend is consistent with recent simulation benchmarks such as RoboLab. The policy behaviors in the real world:

Introducing HO-Cap: A Capture System and Dataset for 3D Reconstruction and Pose Tracking of Hand-Object Interaction. We built a multi-camera system and a semi-automatic method for annotating the shape and pose of hands and objects Project page:

Big day in class today! With Luis Felipe Casas, we brought 14 SO-101 arms for students to assemble and take home for projects UT Dallas. Excited to see what they create by semester’s end! 🤖

I was preparing a video to introduce our lab Intelligent Robotics and Vision Lab @ UTDallas for a meeting. Happy to share the video here! We are looking forward to collaborating with both academia and industry. Please feel free to reach out

We can replay the HO-Cap data in #IsaacSim No physics enabled. It is setting the poses of hands and objects according to our annotations

Dobot Dobot Robotics visited us today and showcased their 6-legged robot dog at The Jonsson School at UT Dallas!

We Luis Felipe Casas have successfully deployed an in-hand manipulation policy to the LEAP hand following RL training in MuJoCo Playground. Sim-to-real isn’t as straightforward as it seems. Policy design and real-world tuning matter. Working on improving the policy further.