Jetson Thor day 1: local VLM + RealSense 3D vision describes scenes every 350ms
chrismatthieu · x · 2026-09-03
The author's day-1 build with NVIDIA Jetson Thor: a fully local robot perception stack combining Intel RealSense RGB+depth with a local 4B Qwen VLM that describes what it sees roughly every 350ms, plus real-time distance estimates in left/center/right directions and center-mass range — all on-device with no cloud.
The takeaway: pixels tell AI what the world looks like, depth tells robots where and how it works. Next step is mounting the whole setup on an actual robot.
Related event: Devs Test NVIDIA Jetson Thor Running VLMs Locally(2 posts)→
More from Embodied
- Researchers turn to DSRL to improve BC diffusion policies via latent-space RL — DominiqueCAPaul · 2026-09-03
- HT Robotics' ~$5k Mini Pi humanoid faces price pressure from $399 MiniDuck — philfung · 2026-09-03
- Waymo Adds 128 Robotaxis in Texas, Growing Fleet to 960 Vehicles — reed · 2026-09-03
- Yacine says RNNs learn sysid anyway and vision policies aren't that hard — yacineMTB · 2026-09-03
- Robotics' demo-to-reality gap will dwarf LLMs', argues new post listing 14 challenges — Ben_Reinhardt · 2026-09-03
- Video of an 80-year-old with a cane tailed by a robot goes viral — TansuYegen · 2026-09-03