This is a small tool written in Python 3 to visualize and process the Raspberry Pi Cameras H264 motion vectors.

Option 1: Record the motion vectors and save it to a file on the Raspberry Pi to re-play it later:

Run raspivid -o /dev/null -x vect.txt -t 0 on the Raspberry Pi. The -t 0 option means that it will record until you quit, -t 5 would quit after five seconds. If you also want to save the corresponding video, you can specifiy the out file -o vid.avi.
Copy the file to the computer where you want to re-play it. If you want to re-play it on the Raspberry Pi just leave it there. You can copy for example with scp.
Assuming your file is named vect.txt run cat vect.txt | python proc.py. The file should be re-played now.

Option 2: Transfer the motion vectors via TCP to a client that runs the processing:

This is especially handy if you are working on a Raspberry Pi zero with no screen attached.

Run raspivid -l -o /dev/null -x tcp://0.0.0.0:1234 -t 0 on the Raspberry Pi. This runs raspivid and makes it listen on TCP port 1234 on the local IP address. (hint: as soon as you connect to the port and disconnect afterwards, raspivid will terminate. You can run it in a loop on the Raspberry Pi with while true; do raspivid -l -o /dev/null -x tcp://0.0.0.0:1234 -t 0; done).
Run nc raspberrypi.local 1234 | python proc.py on the computer where you want to visualize the optical flow. If you have Avahi running then raspberrypi.local should work, otherwise replace it with the IP address of your Raspberry Pi.

This is useful if you want to process the motion vectors directly on the Raspberry Pi.