Questions about Jetson TK1 for use in a commercial project

Hi All,
This is my first post to the community.
I was considering buying the TK1 for my work(described in point 1) but I had a few queries regarding the same. Hope someone can help me out or advice:

  1. Is the TK1 good/stable enough to create/sell a system for commercial usage? The idea is to integrate the project OpenANPR for number plate recognition and place Jetson boards near entry/exit gates of a building. Also maybe use IO pins for entry/exit gate control

  2. How can one debug easily(and productively) the OpenCV code on the Jetson? Use a JTAG? if yes then has someone tried it and any links to tutorials and suggested JTAG kits? If there is other ways to do the same then kindly recommend that too.

  3. How easy it is to port an openCV code with CUDA support onto JEtson Tk1? is it simply equal to compiling it for Jetson and it will run out of box? or some considerable work needs to be done?

  4. Is there any news about when Tegra X1 based boards will be out? They look like more powerful for image recognition

  5. Any comments about the idea I proposed in Point 1?

  6. Suggestions for any other board(from anyother company) for my work?

  7. Do I need to consider the Tk1 Pro version for my work?

Hoping for any pointers

BR
general_heat

Bump…Any comments from the stalwarts at the forum.
Thanks in advance

I don’t much about the topics, but I’m guessing something here.

  1. I think the Jetson is a development platform only and it’s not probably meant for actual embedded products. The idea probably is that you can quickly test and prototype your idea on the Jetson and then design the exact product you need based on the same SOC.

  2. JTAG is typically meant for a lower level debugging (like porting a bootloader or kernel). For OpenCV you can use normal application level debugging tools (GDB, debug prints). I don’t know if there are OpenCV specific tools to measure performance or bottlenecks. You should be able to even develop your application on a Linux PC with NVIDIA GPU (and CUDA) and then just recompile it to test the performance on the Jetson. Of course there are differences but if you take them into account, I think your should be able to run same code on both platforms.

  3. I think Tegra version of the OpenCV already supports CUDA to some extent. For more info, check: http://elinux.org/Jetson/Installing_OpenCV

  4. I haven’t seen any schedules for generic purpose X1 boards. X1 certainly is faster than K1 but every now and then there’s something new and faster :)

  5. Sounds doable and Tegra K1 is supposed to be good in image processing. I don’t have personal experience on that (yet). But as said, it might not be possible to use Jetsons for it (but I don’t know about the licensing).

I recommend that you buy one Jetson (they are not that expensive) and try it out. If the board seems to be suitable for your needs you may want to contact NVIDIA’s HW parters if they have production ready boards (I haven’t heard of any yet though):
https://developer.nvidia.com/tegra-hardware-sales-inquiries

Surely TK1 in stable enough to make a commercial product with it as many people (including myself) have done that already :)