Ai deck camera - upside down image streaming and artifacts

Discussions about the AI-deck
Post Reply
Posts: 4
Joined: Tue Oct 06, 2020 7:43 pm

Ai deck camera - upside down image streaming and artifacts

Post by lorenzolamberti »


I am having some issues in fixing the camera image stream when i load my code with jtag (wifi_jpeg_streamer). The code comes from a clean github pull, so it is the same code available for everyone and behaves differently with respect to the binary pre-loaded in the AI deck

I am using the latest AI-deck, with the grayscale camera.

The image is upside down, even if in the code we should set a register value to rotate it.
Plus, the image goes black if the camera is not pointing to a very bright light (as a lamp, or window)
Finally, there are some artifacts that i am not understanding:
1. in the bottom part of the image only, the white pixels become Black if thelight is too strong. This happens just for a single row of bits at the bottom.
2. There are some squares that occasionally pop up in the image, which seem jpeg encoding artifacts.

I am attaching an image that shows both of these problems.

Most importantly, all of these issues are not happening with the pre-loaded binary, but just with the git code.
any toughts? could you doublecheck the code that is on git?

I assume that it is just a matter of setting registers of the camera for the sensitivity..
immagine_2021-03-18_174905.png (40.04 KiB) Viewed 286 times
immagine_2021-03-18_174835.png (36 KiB) Viewed 286 times
Posts: 749
Joined: Fri Jul 06, 2018 11:13 am

Re: Ai deck camera - upside down image streaming and artifacts

Post by kimberly »


Actually, we are not really sure what is going on! But it is interesting that the version that came on the gap8 worked better. The only difference I know of is that the original wifi version is based on SDK 3.5 and this one on 3.8 (which have some registries added as a fix), other than that there are no differences...

I made an issue on that on the AIdeck repo to investigate it further: Thanks for reporting it!
Post Reply