eereid is a minimalist Python library, allowing users to train and evaluate re-identification models with ease.
Built on tensorflow and developed as part of a research project at TU Dortmund University, the library aims to be a light-weight alternative to commonly used libraries for deep learning and re-identification in particular. The library allows you to easily use your own datasets and model. In doing so, you can tweak training parameters and evaluate your models in a breeze, even if you are not particularly familiar with deep learning.
On your preferred OS, make sure you are using python 3.8 or 3.9: https://www.python.org/downloads/release/python-380/
Open up your preferred directory and clone this repository:
git clone https://github.com/psorus/eereid.git
Navigate to the installation directory and run install.sh in a terminal to install the required modules
sh install.sh
And that's it - With only five required moldues installed, you are now good to go!
To see if the installation went well, you can navigate to the tests folder and run main.py. This will train and test a simple CNN on MNIST and you will be provided with Ranked Accuracy and mAP for evaluation purposes. In main.py, you can now tinker with training and testing parameters such as:
Navigate to the tests folder and open create_datasets.py. Depending on the labeling of your data, you can add the path to your dataset here and simply import it by running create_datasets.py, which will generate an npz-file out of your dataset.
TBD