Closed kazucmpt closed 5 years ago
You can choose b and t by yourself. The important thing is to normalize optical flow images into between 0 and 1.
We want to reproduce your experimental result. We know we can choose b and t by ourselves as you said but we would like to know your parameter values.
Sure, we use default parameters as dense flow.
Which branch of dense flow did you use, master or opencv3-1?
The default values of the parameters are different.
I used the master branch.
kazucmpt notifications@github.com于2018年10月2日 周二下午4:21写道:
Which branch of dense flow did you use, master or opencv3-1?
The default values of the parameters are different.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/hyf015/egocentric-gaze-prediction/issues/8#issuecomment-426174220, or mute the thread https://github.com/notifications/unsubscribe-auth/AVSGeMAGVaifnlt-W_PggaERj-LNKP8Pks5ugxQLgaJpZM4W-Edh .
According to extract_flow_gpu.cpp, the default values of the parameters of the master branch are t=0 and b=15. But in README.md of the master branch of dense flow,
Usage ./extract_gpu -f test.avi -x tmp/flow_x -y tmp/flow_y -i tmp/image -b 20 -t 1 -d 0 -s 1 -o dir
You said you used the defult values. That means t=0 and b=15, right?
Thank you.
Sorry for the confusion. I mean b is 20 and t is 1.
kazucmpt notifications@github.com于2018年10月2日 周二下午4:30写道:
According to extract_flow_gpu.cpp, the default values of the parameters of the master branch are t=0 and b=15. But in README.md of the master branch of dense flow,
Usage ./extract_gpu -f test.avi -x tmp/flow_x -y tmp/flow_y -i tmp/image -b 20 -t 1 -d 0 -s 1 -o dir
You said you used the defult values. That means t=0 and b=15, right?
Thank you.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/hyf015/egocentric-gaze-prediction/issues/8#issuecomment-426176288, or mute the thread https://github.com/notifications/unsubscribe-auth/AVSGeFpNzPYX6dfBRWi2YK5eP58bSn10ks5ugxYOgaJpZM4W-Edh .
Thank you very much.
How many hours did you need to calculate optical flow by using GPU? We had started to calculate the optical flow of GTEA dataset plus by using dense flow a few days ago. But the calculation is not finished yet. We use GTX-1070.
It seems that it takes more than 3 hours in each avi file.
How about you? Did it take so many hours?
Yes, it took a few days I remember...
Thank you very much. Your answers always help us.
To generate input image files, you used dense_flow. What were your parameters b and t?