Deal with double frames in animation, detect scenes, some small improvements.#120
Deal with double frames in animation, detect scenes, some small improvements.#120edgarsi wants to merge 9 commits intobaowenbo:masterfrom
Conversation
ffmpeg select(n indexes from 0, Colab arguments from 1.
Claymation often have models moving only every second frame. Interpolating several frames apart can smooth out even those. Often it's askign too much however, since multiple frames apart is not what the net is trained on. Only implemented for framerate increases 2x the original.
AlphaGit
left a comment
AlphaGit
left a comment
There was a problem hiding this comment.
Nice work!
Consider updating the version number, so that whenever people are reporting issues with the Colab file, they can tell easily if it's the latest version.
Also consider adding yourself to the credits, if you are ok with people reaching out to you. :)
I see you did a major work on that colab_interpolate file -- thank you very much for that! It was in need of some love.
And you added two new interpolation algorithms, wow! <3
| "\n", | ||
| "#@markdown ## Frame mixer\n", | ||
| "#@markdown Normally we interpolate between two adjacent frames. You can try to experiment with other mixers. Claymation mixer can give smoother outputs when animation only moves every second frame at times.\n", | ||
| "FRAME_MIXER = \"normal\" #@param [\"normal\", \"claymation\"] {type:\"string\"}\n", |
There was a problem hiding this comment.
I think greased is missing here:
| "FRAME_MIXER = \"normal\" #@param [\"normal\", \"claymation\"] {type:\"string\"}\n", | |
| "FRAME_MIXER = \"normal\" #@param [\"normal\", \"claymation\", \"greased\"] {type:\"string\"}\n", |
There was a problem hiding this comment.
I found greased to be useless, its results disappointing. Didn't want to pollute the colab API with that. It may be best to remove it completely, but the maintenance burden is small, so left it in.
There was a problem hiding this comment.
Sounds good to me! I'm glad it wasn't a mistake.
|
I ran this in Google Colab and it seems like it is slower (~4.75s/it) than the original (~3s/frame). Any possible reason why? Edit: I think I'll do further testing in the future, it might be because I got a slower GPU while testing. |
You can use nvidia-smi to check which GPU you got, I think in the Colab_DAIN.ipynb, there is a cell in the beginning that has |
|
Is there a reason this hasn't been merged? |
|
I have incorporated these changes into my fork of this repository. |
5 participants
No description provided.