Open DevinBayly opened 2 years ago
created a version of the simplest pass that end's after calculating centroids on the blobs, it's currently running
It also seems like for creating data that can be used in the "OnlyDetects" pass we could very well be processing an individual video using multiple cpu cores by leveraging the seek
method to proceed to a particular frame.
this frame calculation could be something like number of video frames/cores = frames_per_cpu
and then create a processor for each with an argument to skip ahead some number of frames potentially
It would just be important to not export to the same file at the end.
Following this approach and running on 16 cores the estimate has decreased for the DetectOnly
pass to be about 30 mins until finish. Each of the 16 cpu are performing something like 20 iterations per second.
at this commit the speed of the processing is something like ~6x the speed of the video.
https://github.com/DevinBayly/thermal_imaging/commit/cc397e71ee2181681a9564abd3f3f8667a627ad3