I’ve just added an Xavier to my open source four wheel drive robot. For developers, the first Xavier is $1300. I’m really interested in pure camera based navigation and the Xavier is perfect for developing something like that. I’m going to start with using a stereo camera to follow hiking trails and go from there. The latest on the robot is here if anyone is curious:

https://imgur.com/a/GqXD2Zj

Are you having any trouble with USB and that ps4 stereo camera? I got a Xavier to process video from intel realsense 415 stereo cameras and am having a lot of USB problems with dropped frames and frames missing data. The cameras work fine on x86. The entire Xavier software stack feels very early beta quality.Not sure if I should abandon the Xavier, or try a PCIe USB card or what.

I have not used the USB camera extensively. I installed the experimental drivers (https://github.com/ps4eye/ps4eye/blob/master/README.md) and got it working, then left it displaying for a couple of days. It never froze, but I wasn’t watching for dropped frames. Now I’ve mounted the Xavier on the robot, but haven’t started developing the application with it.

I definitely remember that when you download the system image for the Xavier, they said it was a beta version or something like that. I haven’t seen any issues in my limited use, but I did get the sense it wasn’t fully baked. Hopefully your issues get worked out soon. It would be interesting to see how long the TX2 development took to stabilize.

I highly highly recommend using a PCIe USB card, especially if you are using multiple cameras.

Thank you so much, I got a PCIe USB card and it fixed all my problems.

Wow. I had a lot of fun with the JTKs when they first came out. I set up a small render farm, and had a blast with them. But at a $2500 price point, I can't see broke, indebted, just-out-of-college me getting the opportunity to do that with this new dev kit, and I think that's kind of a shame. With that said, this is pretty cool, and I totally want one

My team managed to do pretty cool realtime deep neural net computer vision demos on Jetson TX2, but this beast is an order of magnitude faster and energy efficient, which seems almost unbelievable! We're definitely getting a bunch of those and sticking them to our robot.

This fills a need for a general development platform for Nvidia's ARM modules. You could get dev boards for these modules, but the dev boards were (mostly) not suitable for prototyping or products. I think a lot of companies would just default to x86 platforms because of the ease of sourcing small form factor boards that have plenty of connectivity. However, for the right application the performance per watt and form factor is a winner. I guess something that needs low power and a lot of AI/Image processing in a small brick. Autonomous vehicles?

Bingo. NVIDIA made a fairly big stink about ASIL qualification efforts for Xavier.

Is there a big market for this?

In five years, this style of device will be huge in robotics. NVIDIA is getting the industry hooked now.

Cars seem like a pretty big market.