We need to soon investigate the problems we spotted while using ZFP in the whole demo to save bandwidth for transmitting depth images, which essentially make wrist key points disappear.
At this point, we have only tested that w/o ZFP the whole demo works smoothly, whereas w/ ZFP we see missing key points (due to the flawed depth, presumably).
We might want to test ZFP-0.5.5 (released recently) that provides lossless compression, in case we'll discover that the quantization level is the problem.
We need to soon investigate the problems we spotted while using ZFP in the whole demo to save bandwidth for transmitting depth images, which essentially make wrist key points disappear.
At this point, we have only tested that w/o ZFP the whole demo works smoothly, whereas w/ ZFP we see missing key points (due to the flawed depth, presumably).
cc @vvasco @barbalberto