Interpolate entire sequence

[Jerchongkong]

when I try to interpolate a sequence (1 minute of video) it only interpolates the first frame and nothing else.
example: if the interpolation is at 0.25 it only interpolates the first 4 frames and the process stops.

what's the problem in this?

[Jerchongkong]

It seems that nobody knows the solution, or perhaps it is a question too stupid to be taken in the field. I don't know if it has to do with ubuntu 18.04 and cuda 9

[vincentleon]

It interpolates between the two frames defined in arguments_strFirst and arguments_strSecond. If you want to do for a whole sequence, you have to change the code, add a "for" loop, and interpolate between each pair of frames.

[Jerchongkong]

It interpolates between the two frames defined in arguments_strFirst and arguments_strSecond. If you want to do for a whole sequence, you have to change the code, add a "for" loop, and interpolate between each pair of frames.

I'm sorry, I'm a complete noob on this :v

How am I supposed to do that? Do you think you can send the edited code to interpolate the entire sequence of images?

[fathomson]

How am I supposed to do that? Do you think you can send the edited code to interpolate the entire sequence of images?

I ran into the same, and solved it with a for loop:

frames = sorted(os.listdir(os.path.join(MB_Other_DATA, dir)))
for i in range(0, len(frames)-1):
    arguments_strFirst = os.path.join(MB_Other_DATA, dir, frames[i])
    arguments_strSecond = os.path.join(MB_Other_DATA, dir, frames[i+1])

Note that at the end of the script you also need to modify the count in order to save time images, so:

count = i * (numFrames+1)

[Jerchongkong]

How am I supposed to do that? Do you think you can send the edited code to interpolate the entire sequence of images?

I ran into the same, and solved it with a for loop:

frames = sorted(os.listdir(os.path.join(MB_Other_DATA, dir)))
for i in range(0, len(frames)-1):
    arguments_strFirst = os.path.join(MB_Other_DATA, dir, frames[i])
    arguments_strSecond = os.path.join(MB_Other_DATA, dir, frames[i+1])

Note that at the end of the script you also need to modify the count in order to save time images, so:

count = i * (numFrames+1)

It seems that it does not work, it only interpolates the first 3 frames.

I was able to run DAIN on google colab, giving the same result. For anything I leave the link. Maybe you can solve the problem.

https://colab.research.google.com/drive/1_Yf6IY3Nbc-Jtn91qx0wcwWOVe0Zkbgl?authuser=2

[WMCh]

How am I supposed to do that? Do you think you can send the edited code to interpolate the entire sequence of images?

I ran into the same, and solved it with a for loop:

frames = sorted(os.listdir(os.path.join(MB_Other_DATA, dir)))
for i in range(0, len(frames)-1):
    arguments_strFirst = os.path.join(MB_Other_DATA, dir, frames[i])
    arguments_strSecond = os.path.join(MB_Other_DATA, dir, frames[i+1])

Note that at the end of the script you also need to modify the count in order to save time images, so:

count = i * (numFrames+1)

It seems that it does not work, it only interpolates the first 3 frames.

I was able to run DAIN on google colab, giving the same result. For anything I leave the link. Maybe you can solve the problem.

https://colab.research.google.com/drive/1_Yf6IY3Nbc-Jtn91qx0wcwWOVe0Zkbgl?authuser=2

I am not sure if I understand your needs. Demo_MiddleBury.py is used for 1-frame interpolation and demo_MiddleBury_slowmotion.py is used for multi-frame interpolation between two ajacent frames by setting time_step. nframe=(1/time_step)-1. Maybe you can refer to @fathomson's comments and modify the corresponding script to suit your needs.