The DeepStack documentation now has a FAQ page with answers to common questions on DeepStack and each of the available APIs. Visit the link below for any questions you may have.
If your question is not on the page, reply to this post and it will be added to the FAQ page along with i the answer.
Hello Moses. I was trying to find the minimum requirements for the GPU version, maybe that would be useful to add.
It seems that it only work with Nvidia?
I just found a very cheap ASUS GeForce GT 710 but it has only 1GB GDDR5, would that work? If you could specify the minimum requirements for the graphic card it would be very helpful.
I want to be able to define the minimum cost of an installation for the AI tool with Deepstack and there is a big range of GPU prices. Maybe buying a GPU to install on an older PC would be more convenient than buying a new PC to use the CPU version.
Thank you
I suggest a simple set of instructions for upgrading. My particular interest is the native Windows CPU version. Is it a recommended practice for every upgrade to uninstall the previous version and do a fresh install?
@mrpie We recommend the use of a graphics card with a minimum of 4GB memory to run DeepStack.
This will be added to the FAQ.
You don’t need to uninstall the existing version. Simply
If I’m running Deepstack to interpret images for multiple cameras, can I submit them all to the same Deepstack container or should I run a container for each camera?
I’m only thinking to ask this because I looked inside the container at the contents of the Redis and it appears to contain past detection results, so I’m wondering if you confuse Deepstack by co-mingling images from different cameras.
@sprior You can submit images from any number of cameras to a single container instance of DeepStack. You don’t need one instance of DeepStack per camera.
However, do take note that the number of frames you can process per instance of DeepStack is limited by the capacity of your machine, which means the FPS processing rate for each camera diminishes as you connect more camera sources.
When I was poking around in Redis in the container I saw a bunch of keys which I’m guessing weren’t cleaned up from previous requests and didn’t have a TTL set. Would it be a good idea to add a TTL to response keys so that they’ll be cleaned up automatically after some timeout period? Otherwise it seems like a slow/small resource leak in the container.