I suggest create new python virtual environment and work in it's to you. Installiations can
disorder your python packages.
For create:
`python -m venv new_env`
For activate:
`.\new_env\Scripts\activate`
There are labels as coco json format on our hands. For YOLO training, we must convert it to yolo format.
We can use `dataset_manager.py` for this operation. The python file provides some convenience to us.
* Convert COCO to YOLO format
* Counting the number of labels for each class
* Masking object that at chosen class on images and deleting labels in .txt files
* Convert source of video to images
You can follow the `detect_and_track.ipynb` file for training for detection and adding track feature
on structure.
You can access models and ouput videos with this Link