You probably don't want to fork it 😜
- Scrap videos using
youtube-dl
orinstaloader
. You can usetools/scrapping.py
to automate this step. videos_add.py
: Copy and convert videos into a folder in (seeconfig.py
for list of options) Then extract for each video:- Scenes - using
ffmpeg / ffprobe
- Frames - as images
- Features for each frame - using the
classifier
- Predictions for each frame (labels) - using the
classifier
- Scenes - using
videos_process.py
: Builds (or load) single files concatenating all:- Scenes
- Frames
- Features (Reducing dimensions using PCA)
- Predictions It also create a file holding the graph connecting all features.
videos_test.py
: Runs different test such as finding closest images and finding path between 2 framesstartend_images_encode.py
is used to encode some images to start and end the animation.startend_images_getpath.py
is used to test the start and end image (this is a quick last minute fix that needs to be removed)build_annoy_index.py
encode pregenerated lines of poetry usingspacy
and builds an index of those lines usingannoy
api.py
starts a local api which can be called togetpath
between 2 images orGetSimilarVideos
P 3.6. Need to add version for all packages
- keras
- tensorflow
- igraph
- numpy
- scipy
- pickle
- matplotlib
- tqdm
- sklearn
- rasterfairy
- moviepy
- PIL ??
- imageio ??
- instaloader
- annoy
- spacy + en-core-web-sm / en_vectors_web_lg
- termcolor
- colorama
- youtube-dl
- ffmpeg / ffprobe
import yt video etc... db?