- Needs ffmpeg installed.
- Ensure Gstreamer is installed on Ubuntu / Klite codec is installed on Windows
- Requires Cuda 10.1 and Cudnn Installed to run tensorflow with GPU. Follow installation shown in tensorflow-gpu installation guide.
To use this module,
- Git clone this project.
- Navigate to base folder and run installation with: ipython installation.py This might take a while as it downloads the ZQPei/deep_sort_pytorch module from https://github.com/ZQPei/deep_sort_pytorch.git to the base directory, and download the neede pretrained weights for yolov3.
- Create a python virtual environment and install the requirements needed listed in "requirements.txt" with the following code depending on the version:
virtualenv --no-site-packages --distribute .env &&\
source .env/bin/activate &&\
pip install -r requirements.txt
virtualenv .env && source .env/bin/activate && pip install -r requirements.txt
You can check your virtualenv version with the following line on command prompt:
virtualenv --version
To perform prediction on input videos, you can choose to open the UI by navigating to UI folder and running:
python 0_mainUI.py
or choose to run the prediction file manually after setting appropriate variables with:
python run_prediction.py