Closed Cioscos closed 3 years ago
It only works with single frames. About RAM issues, this version loads all the frames into GPU RAM and some Numpy manipulations also loads it entirely into RAM.
You may try my fork, which loads and process frame by frame. Two considerations to take in account is the multiprocessing with Ray, as it consumes around 2GiB of RAM per task-cpu-core on a 720p frame. The other consideration is that the second step will consume far more memory as I were unable to optimize flowNN process and my per-batch approach to remediate that is experimental.
It only works with single frames. About RAM issues, this version loads all the frames into GPU RAM and some Numpy manipulations also loads it entirely into RAM.
You may try my fork, which loads and process frame by frame. Two considerations to take in account is the multiprocessing with Ray, as it consumes around 2GiB of RAM per task-cpu-core on a 720p frame. The other consideration is that the second step will consume far more memory as I were unable to optimize flowNN process and my per-batch approach to remediate that is experimental.
Ok, thx I will try ur repo
Hi! Can I use it directly with a video file? Putting the corresponding mask video on another folder? Or i need single frames?
I tried with single frames but seems that ram finishes very fast. With colab pro is it usable? In general this application can manage folder with 2600 files?