DimensionFour
DimensionFour copied to clipboard
An open source Video Synopsis tool
DimensionFour
Installation
Make sure you have have the pipenv package manager installed
pip3 install pipenv
Install dependencies using pipenv
pipenv shell
pipenv install
Running the code
Command line usage
- Enter the environment using:
pipenv shell
- Preprocess your video files:
- Individally as:
python3 -m dimensionfour.preprocess --input dataset/1.mp4
python3 -m dimensionfour.preprocess --input dataset/2.mp4
...
- Or in bulk as:
python3 -m dimensionfour.preprocess --input dataset/*
The first time you run dimensionfour.preprocess, it will download a pretrained yolov3 model from dropbox.
Preprocessing will output a file called <filename>.d4artifact.zip into the current directory which contains metadata about your video
- Assemble your
.d4artifact.zipfiles into a final video:
python3 -m dimensionfour.assemble --input *.d4artifact.zip --output summary.avi
This will output a summarized video of all d4artifacts in the current directory as summary.avi
Options
You can filter the summarized video to only include certain object classes using the --filter option:
Default: No Filter
python3 -m dimensionfour.assemble --input *.d4artifact.zip --output summary.avi --filter person
You can set the fps of the output using the --fps option:
Default: 30
python3 -m dimensionfour.assemble --input *.d4artifact.zip --output summary.avi --fps 10