This tool has two basic mode. At first you can generate training data for manual labeling in cvat here is the relevant fork.
The other mode is for planned capturing the Videostream.
I recommend using a virtual Env or a conda/miniconda Environment for executing this scripts.
Please install all necessary packages with
pip3 install -r requirements.txt
If you want to do it the manual way you need:
- opencv
- python-decouple
- numpy
- python-dateutil
For the camera access and automatic upload to the azure file storage you need to provide credentials, usernames and tokens
like in this example file. Run cp env.example .env and place your secrets into .env
If you don't want to use the azure upload function, the sas_token and filepath is not necessary for local capturing and storing videos.
Just run
python3 main.py
and a video will be captured with a live view. To finish the capturing just press q on your keyboard.
Run
python main.py -r -rd 2 -vd 1
This runs the script 2 minutes and every 1 minute a new video file will be created
If you just want some random images during the script runtime, just run
python3 main.py -i
This will save every minute one image from the stream. This is perfect for generating a huge amount of training data for labeling.
Run python3 main.py -u uploads the video file after capturing finished directly to azure file storage.