It looks like you're new here. If you want to get involved, click one of these buttons!Sign In
@Tome_at_Intel: I think that there is great potential for using the NCS for computer vision applications, especially on low-power platforms such as the Raspberry Pi. Since the Myriad processors seem to directly aim at this use case, I am surprised that it is actually not possible to use this functionality with any of the devices currently available to non-OEMs:
Please correct me if I'm wrong, but it seems that all of these devices only work together with the NC SDK, which currently only supports Caffe and TensorFlow.
I've been tinkering with the NCS for a while now, however, my intention was to use it for tasks such as image filtering, feature detection, stereo matching and visual odometry. Unfortunately, at the moment it seems not to be suitable for this type of applications.
I think that the NCS is a good product and that it has great potential for computer vision tasks in mobile robots, quadcopters and dozens of other vision-based applications. However, at the moment the only thing you can do with it is image classification. It's probably a nice experiment to get a deep convolutional neural network running on a Raspberry Pi, and maybe it's targeting educational purposes, but I really can't see any meaningful real-life applications for the NCS in combination with the NC SDK in it's current form. So I wonder what the long-term goal is.
While reading through this forum I found many discussions addressing computer vision and the MDK, so there seem to be many other developers who are interested in this subject, however I could not find all the answers I was looking for.
So I would like to ask the following questions:
Everyone who is interested in computer vision is welcome to comment.