Error on "Load driving video and source image" section of colab

[MaalikN]

---

RuntimeError Traceback (most recent call last)
in ()
9
10 source_image = imageio.imread('/content/gdrive/My Drive/first-order-motion-model/ev2.jpg')
---> 11 driving_video = imageio.mimread('/content/gdrive/My Drive/first-order-motion-model/brnew.mp4')
12
13

/usr/local/lib/python3.6/dist-packages/imageio/core/functions.py in mimread(uri, format, memtest, **kwargs)
305 ims[:] = [] # clear to free the memory
306 raise RuntimeError(
--> 307 "imageio.mimread() has read over 256 MiB of "
308 "image data.\nStopped to avoid memory problems."
309 " Use imageio.get_reader() or memtest=False."

RuntimeError: imageio.mimread() has read over 256 MiB of image data.
Stopped to avoid memory problems. Use imageio.get_reader() or memtest=False.

[MaalikN]

driving_video = imageio.mimread('/content/gdrive/My Drive/first-order-motion-model/brnew.mp4', memtest=False)
seems to work, is there a reason not to add memtest=False?
The also seems to be slowed down, why is this?

[AliaksandrSiarohin]

If size of the video is too large it will cause notebook to crash.
Which cell is slowed down?

[MaalikN]

The result from "Load driving video and source image" and the end result of the animated image. The original video is around 29 seconds, the end result is 44.

[AliaksandrSiarohin]

You should set fps. Imageio.mimsave(...,fps=...)

[MaalikN]

I ended up cloning the git via anaconda and running it locally. Now it works great! Is there a tutorial on how to get face-swap to work?

[animemoeus]

Same problem.
From the documentation here, I just tried to disable memtest and its work for me.

[alexjavorski]

Did you disable the memtest with memtest=False?
If so where and how did you use this line?
@animemoeus

[animemoeus]

Did you disable the memtest with memtest=False?
If so where and how did you use this line?
@animemoeus

[alexjavorski]

Really? Then I mist be doing something else wrong. Thanks

[kcrank1]

@MaalikN did you solve your video slowing down issue? I was able to bypass the error with memtest=False but I still get a 23 sec video rather than 15 sec as my original video is

[MaalikN]

@kcrank1 I ended up just setting it up on my own computer to run on my hardware instead of google colab. When running on my own gpu it doesn't slow the video down. A solution that I used previously though, was to compare the length of the driving video and the length of the produced video to find what percent the produced one was slowed down by. Then you can speed it up by the exact amount in most editing software.

[kcrank1]

@MaalikN that's a great workaround. What I ended up doing to get around the problem was noticing that the output video was 20fps, if I edited my input video to be 20fps going in it preserved the speed.

[Wei-En-Chen]

How can I solve this problem?
The video length of my training data and test data is about 10 seconds. There are 2000 training data and 500 test data.
dataset is mp4 file, picture length:1920 picture width:1080 fps:60