Dockerize the application that required opencv, tensoflow, ffmpeg with GPU

  docker, ffmpeg, kubernetes, machine-learning, opencv

I have a pretty complicated video analysis application. The application takes the video as input, performs the image classification by Inception that required GPU to accelerate, then the application outputs several clips from the original videos by Ffmpeg accelerated by CuDa with GPU.

Basically, I need to install anaconda with python 3 and many required packages. Beside that I need build opencv, ffmpeg from code also.
Now, to auto scale, I ‘d like to dockerize this application.

Do you have any ideas or have any suggestions for me? Is it possible?
Thanks and regards.

Source: Docker Questions