I have Implemented a super light Machine vision algo on desktop using an Nvidia card but its suffering from high latency prob due to sloppy pipelines and lack of unified memory.
Was planning on implementing on PYNQ but I’ve noticed most if not all PYNQ devices are all MPSoM, with ARMs. Ideally id want to utilize only programable logic because I’m looking for super low latency. I’m questioning how much the ARM cores are being used and for what? I guess looking at SoC block diagram might answer my question but for me this isn’t that apparent. Some devices in question are the KV260(shown) and PYNQ-Z1(and Z2).