kabir
@mhkabirr
Followers
59
Following
660
Media
1
Statuses
24
Flying robots to save the world
Mountain View, CA
Joined August 2011
Massive dataset but egocentric video is not going to get us dexterous manipulation policies. Egocentric data provides high-level semantic scene and task understanding (which frontier VLMs already generally provide). What is needed is fine sub-mm-level finger pose & force data.
today, we’re open sourcing the largest egocentric dataset in history. - 10,000 hours - 2,153 factory workers - 1,080,000,000 frames the era of data scaling in robotics is here. (thread)
18
14
199
Symmetric DRAM layouts save so much time. Even if the controller pinout is not symmetrical (which it rarely is) there are still so many swap options that it makes sense to mirror most of the escape routing
14
46
629
Now everyone's doing it??? 💅 "While the High Bandwidth Memory (HBM) subsystems were the most sensitive component, they only began showing irregularities after a cumulative dose of 2 krad(Si) — nearly three times the expected (shielded) five year mission dose of 750 rad(Si). No
Our TPUs are headed to space! Inspired by our history of moonshots, from quantum computing to autonomous driving, Project Suncatcher is exploring how we could one day build scalable ML compute systems in space, harnessing more of the sun’s power (which emits more power than 100
0
0
0
tfw you start recognizing hostnames of specific nodes on the cluster
5
3
171
Love seeing high automation lines like this one - we'll need a lot more to profitably reshore manufacturing in the near term we are so back
Late last month, our G1_Dallas factory set a new production record: 14 MW of solar panels in a single day. Do the math: that's a 5.1 GW annualized pace. Domestic solar. It's the path to American power. Thank you to our G1 workers!
0
0
2
It’s kind of scary to think that if there were a technology that was extremely useful to humanity to build, but it was extremely capital intensive, and building it incrementally wouldn’t have a smooth gradient of returns/revenue, we might just never build it.
393
149
3K
I've been applying ML to "classic" 3D/imaging vision problems for 15 years, so for 14 years I've been rebutting reviewers who don't like seeing _statistics_ (ew! scary!) applied to classic problems. Here's a snippet from a decade-old rebuttal that, regrettably, still holds up.
10
25
380
Mohammed Kabir from @CorvusRobotics talking about the Corvus One, and their very clever use of ROS+PX4 #ROSWorld 2021
2
4
9
📣 Don't miss Mohammed Kabir, CTO from @CorvusRobotics, on his talk at the #PX4DevSummit ✨ Corvus One: Persistent, Vision-Based Aerial Autonomy with PX4 ✨ 💡 Check out the entire program https://t.co/Vcm8RREFlq
0
3
8
YC S18's @CorvusRobotics has built Level 4 autonomous UAVs to help warehouses track inventory. Corvus' solution involves autonomous drones that can fly unattended for weeks on end, collecting inventory data without any human intervention at all. https://t.co/obfFnm7qN0
0
10
42
This is how you build a robotics startup. @CorvusRobotics just launched fully autonomous (meaning no human piloting or oversight) drones that continuously track inventory in warehouses. Huge, valuable problem, and their waitlist is getting very long. https://t.co/tpyfJJoXmi
2
8
37
Corvus Robotics’ Autonomous Drones Tackle Warehouses https://t.co/W6jUGhJRGv
1
8
25
Very cool fully autonomous warehouse drones from @Corvus_Robotics
3
9
22