optical_flow_gpu_opencv3

Use GPU to extract videos' dense optical flow by opencv3

How to use

Option	Default	Note
f	ex.avi	filename of video
x	img/x	filename prefix of flow x component
y	img/y	filename prefix of flow y component
i	img/i	filename prefix of image
b	20	specify the maximum (px) of optical flow
t	1	specify the optical flow algorithm(default TVL1)
d	0	specify gpu id
s	1	specify the step for frame sampling
h	0	specify the height of saved flows, 0: keep original height
w	0	specify the width of saved flows, 0: keep original width

First of all, you need opencv3 in your environment.(I tested it on my opencv-3.4.0)

pkg-config --modversion opencv
3.x.x

Then just do this to process one video.

./compile.sh
mkdir img

# you can simply use it by
./get_flow_gpu -f=ex.avi

# or you can set some parameters
./get_flow_gpu -f=ex.avi -x=img/x -y=img/y -i=img/i -b=20 -t=1 -d=0 -s=1 -h=0 -w=0

If you have video's folder like this

--video_dir_path
 |--class1
 |--class2
 |--...
 |--classn
   |--classn_video1.avi(.mp4)
   |--classn_video2.avi(.mp4)

You can use video2flow.py to save optical flow image in /some/path/flow_dir_path

python video2flow.py /some/path/video_dir_path /some/path/flow_dir_path