Possibility of intergrate direct into tensorflow workflow?

Based on my experience, the best case of using nvidia opticalflow combined with tensorflow is give it a placeholder and feed the calculated flow into the NN. But is it going to have a performance issue doing this FEED operation? Or is there any existing method to utilize NV optflow directly in tensorflow workflow?

Many thanks!

Hi.
There is a python wrapper available. Can you look into below link and see if this suits your requirement.
https://devblogs.nvidia.com/opencv-optical-flow-algorithms-with-nvidia-turing-gpus/

Thanks.