To get started, install the proper dependencies either via Anaconda or Pip. I recommend Anaconda route for people using a GPU as it configures CUDA toolkit version for you.
# Tensorflow
conda env create -f ALL-IN-ONE-CCTV.yml
conda activate ALL-IN-ONE-CCTV
(TensorFlow 2 packages require a pip version >19.0.)
# TensorFlow
pip install -r requirements.txt
Make sure to use CUDA Toolkit version 10.1 as it is the proper version for the TensorFlow version used in this repository. https://developer.nvidia.com/cuda-10.1-download-archive-update2
Download pre-trained weights file: https://drive.google.com/file/d/1eU_9UjVwhpilSw4pL4curM-Hsu_qBqLA/view?usp=sharing
Copy and paste all folder from your downloads into the this repository.
Huge shoutout goes to hunglc007 and nwojke for creating the backbones of this repository: