Implementing Computer Vision Library
Implementing Computer Vision Library
Hi, is there a way that I can implement computer vision libraries on ai-deck? If so, how should I do that?
Re: Implementing Computer Vision Library
Hi!
No, I don't think that is possible to implement something like opencv on the aideck. Either it has to be build from scratch or you need to use Greenwaves tools in order to make this happen. I would suggest you to make an issue on the GAP8 SDK to ask bout this functionality.
No, I don't think that is possible to implement something like opencv on the aideck. Either it has to be build from scratch or you need to use Greenwaves tools in order to make this happen. I would suggest you to make an issue on the GAP8 SDK to ask bout this functionality.
Re: Implementing Computer Vision Library
Thank you! I will go ask around.
-
- Beginner
- Posts: 7
- Joined: Mon Mar 08, 2021 8:16 pm
- Location: Frankfurt, Germany
- Contact:
Re: Implementing Computer Vision Library
I was looking for something like a CV library as well. I found
https://github.com/liuliu/ccv
which is pure C and very simple to compile.
However, it uses floating point quite a bit which would make it slow, even if I could get it to run (which i couldn't). Further, it obviously cannot take advantage of the parallelization provided by the GAP architecture. I just wanted to use a feature extractor and descriptor like Harris/Brisk or something, but that turns out to be a lot of work. CCV only has SIFT, which is very heavy for something as small as the GAP8.
Another cool feature would be a pixel flow program for the GAP8, because that would allow to extract vectors of movement - usable for stabilizing yaw or height when flying over an object. Has anybody started/found something like that?
https://github.com/liuliu/ccv
which is pure C and very simple to compile.
However, it uses floating point quite a bit which would make it slow, even if I could get it to run (which i couldn't). Further, it obviously cannot take advantage of the parallelization provided by the GAP architecture. I just wanted to use a feature extractor and descriptor like Harris/Brisk or something, but that turns out to be a lot of work. CCV only has SIFT, which is very heavy for something as small as the GAP8.
Another cool feature would be a pixel flow program for the GAP8, because that would allow to extract vectors of movement - usable for stabilizing yaw or height when flying over an object. Has anybody started/found something like that?