temporal-consistency-visualizer: A Python repository from lucidBrot

Usage

Built for python 3.9.2.

If it does not work for your older python version, it might be enough to just comment out all print statements containing =}.

Assumes the frames are image files containing a single number in the filename, which is used for sorting them.

Any frames whose image file names start with weights are ignored.

Direct Usage

python main.py v -i"N:\Temp\videodir" -r400 -o ".\example-outputs"

python main.py visualize --inputdir "N:\Temp\videodir" --row 400 --outputdir ".\example-outputs"

Usage

#: python main.py v -h
usage: main.py visualise [-h] -i INPUTPATH [-o OUTPUTPATH] [-r R] [-n MODELNAME]
                         [-b FIRST_FILENAME] [-e LAST_FILENAME] [-s STRIDE]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUTPATH, --inputdir INPUTPATH
                        Path to the directory containing the numbered frames
  -o OUTPUTPATH, --outputdir OUTPUTPATH
                        Path to the directory where the output should be saved in. (default:
                        directly in ./outputs relative to main.py)
  -r R, --row R         Row to visualize change of. (default: middle row)
  -n MODELNAME, --name MODELNAME
                        Postfix with MODELNAME will be added to generated files.
  -b FIRST_FILENAME, --start FIRST_FILENAME, --begin FIRST_FILENAME
                        Any files before this filename will be ignored.
  -e LAST_FILENAME, --end LAST_FILENAME
                        Any files after this filename will be ignored.
  -s STRIDE, --stride STRIDE
                        How many frames to skip in each step. (default 1)

Custom Quick Usage

Mainly for my own repeated usage for visualizing the results of multiple models:

Specify the BASEPATH variable in the script to a place where directories exist like
```
{basepath}/{modelname}/video{vid}/clip{clip}/
```
Modify the the_dict in forallmodels to map your modelnames to your model's directory. The modelname will be used for further displaying, file naming, and interaction.

python main.py r -v1 -c0 -r200
# runs the visualization of all specified modelname's video1 clip0 on row 200
# specifying row is optional
# synonymous:
python main.py runall --video 1 --clip 0 --row 200

Usage

#: python main.py r -h
usage: main.py runall [-h] -v V -c C [-r R]

optional arguments:
  -h, --help       show this help message and exit
  -v V, --video V  Number of the video
  -c C, --clip C   Number of the clip
  -r R, --row R    Which row. Default is the middle row.

Examples

Ocean 1

# get a video
youtube-dl https://www.youtube.com/watch?v=Mci4m-6WoyM -f bestvideo 
# split it into frames
ffmpeg -i Under\ The\ SEA\ 4K\ Underwater\ Wonders\ +\ Amazing\ Music\ -\ Coral\ Reefs\ \&\ Colorful\ Sea\ Life\ in\ UHD\ 🐟\ 🌊\ 🐠\ \[Mci4m-6WoyM\].webm frame%05d.png
# move the video file out of the way, it would confuse the script
mv Under\ The\ SEA\ 4K\ Underwater\ Wonders\ +\ Amazing\ Music\ -\ Coral\ Reefs\ \&\ Colorful\ Sea\ Life\ in\ UHD\ 🐟\ 🌊\ 🐠\ \[Mci4m-6WoyM\].webm ../video.webm
# take every frame (-s1) from the input directory (-i "N:\Temp\ocean") 
# that lies between(inclusive) the frame00001.png (-b) and the frame00600.png (-e)
# and save
# to the output directory (-o ".\example-outputs") an image file that shows
# how the middle row (omission of -r ROWNUMBER) changed over time
# with filename postfixed with ocean1 (-n)
python main.py v -i "N:\Temp\ocean" -n ocean1 -o ".\example-outputs" -b frame00001.png -e frame00600.png -s 1
main() finished row=540 for naming_postfix='ocean1'

This runs a while.

Ocean 2

Same video, but this time with lower resolution, so it's faster.

# get video
youtube-dl https://www.youtube.com/watch?v=Mci4m-6WoyM -f 'bestvideo[height<=480]'
# this time we limit the threads to avoid the cpu being used at 100%
ffmpeg -threads 2 -i Under\ The\ SEA\ 4K\ Underwater\ Wonders\ +\ Amazing\ Music\ -\ Coral\ Reefs\ \&\ Colorful\ Sea\ Life\ in\ UHD\ 🐟\ 🌊\ 🐠\ \[Mci4m-6WoyM\].webm frame%05d.png
# we could also delete the video now, instead of moving it away
rm Under\ The\ SEA\ 4K\ Underwater\ Wonders\ +\ Amazing\ Music\ -\ Coral\ Reefs\ \&\ Colorful\ Sea\ Life\ in\ UHD\ 🐟\ 🌊\ 🐠\ \[Mci4m-6WoyM\].webm

# run the visualizer again, this time we want a square image.
# according to the output of the youtube-dl command, we've actually got a 854x480 video although we said the height should be at most 480 ??!
# so we make sure the range contains 480 frames I guess.
# I want to start at frame 12519.png and thus stop at frame 12999.png 
# because I'm still using a stride of 1.
python main.py v -i "N:\Temp\ocean2" -n ocean2 -o ".\example-outputs" -b frame12519.png -e frame12999.png -s 1