Waldo Video Preprocessor

Utility script that allows you to convert videos into preprocessed frames. The main use is preprocessing video data for Neural Networks.

Table of contents

• Waldo Video Preprocessor
  • Table of contents
  • Preview
  • How to use
    • With Docker (recommended)
    • Without Docker
  • Options
  • Other notes
    • Remove the container
    • Error messages
    • Bugs
  • Testing

Preview

python3 run.py -s "./data" -f 1

python3 run.py -s "./data" -f 1 --width 128 --height 128 -g

Credits to Homeless. (https://drive.google.com/drive/folders/1-nK8-WthZyB9G1O9MmAgD3h6CQsYkhUr)

How to use

With Docker (recommended)

1. Make sure you have these dependencies installed on your computer:

• Python3 (3.8-3.9): https://www.python.org/downloads/
• Docker: https://docs.docker.com/engine/install/

2. Build the container video with the following command (takes a minute & approx. 1.1GB):

python3 build.py

3. Process videos with the following command:

python3 run.py -s "your_folder_path_with_videos"

When running the script, by default it will create a new folder out/ in your current working directory with all the frames split into separate folders.

Without Docker

There is also the option to run without Docker. However, this can cause errors as you need to manually install and setup opencv-python to work with cv2.VideoCapture().

Make sure to have Python3 installed (see above versions & download link).

1. Set IS_DOCKER to False in variables.py
2. Create virtualenv (optional):

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip

3. Install dependencies:

pip install -r requirements/requirments.txt

4. Process videos with this command (same options as the other 'run.py' command):

python3 main.py -s "your_folder_path_with_videos"

Options

You can view all of the available options with the following command:

python3 run.py -h

The options include:

Arguments	Description	Required	Default	Type
-s --src	The source folder	Yes	-	str
-d --dest	The destination folder	No	"./out"	str
-f --fps	The output target fps	No	10	int
--width	The output frame width. If not provied, the width will not be resized	No	None	int
--height	The output frame height. If not provied, the height will not be resized	No	None	int
--cxmin	The output frame crop min x. If not provided, the frames will not be cropped	No	None	int
--cxmax	The output frame crop max x. If not provided, the frames will not be cropped	No	None	int
--cymin	The output frame crop min y. If not provided, the frames will not be cropped	No	None	int
--cymax	The output frame crop max y. If not provided, the frames will not be cropped	No	None	int
-g --gray	Convert frames to grayscale	No	False	bool
--silent	No console logging	No	False	bool
-t --threads	The numbers of thread to create in order to process the videos (1 file == 1 thread)	No	4	int
-ni --noinput	Prevent the script from asking user input	No	False	bool
-h --help	Show the list of options	No	False	bool

Other notes

Remove the container

To remove the container, simply run:

python3 clean.py

Error messages

You might see this error message in the console:

FATAL: exception not rethrown

For the moment just ignore it.

Bugs

Feel free to open a issue if you encounter any problems. [Open Issue]

Testing

First make sure you have installed the dependencies found in the requirements.txt. (see Without Docker)

Test python formatting:

black --check .

Test python linting:

flake8 .

Python run test suite:

python3 -m unittest discover tests

Here again there is a option to run the tests with or without Docker by toggling the IS_DOCKER boolean in variables.py.