This project is based on the paper: Adaptive Threshold for Better Performance of the Recognition and Re-identification Models.

Please cite this paper:

Bohara, Bharat. “Adaptive Threshold for Better Performance of the Recognition and Re-identification Models.” (2020).

This project can be tested on Labeled Faces in the Wild (LFW), but because of the larger identities, its computationally time consuming. Hence, for testing purpose, I have prepared my own dataset of top highly paid athletes listed on Forbes magazine (2018) using online-dataset-generator. The Athletes dataset can be downloaded from here. Similarly, the model for face-detection and facenet can be downloaded from here.

Clone the project

git clone https://github.com/Varat7v2/adaptive-threshold.git

Install necessary dependencies required for running this project.

pip install -r requirements.txt

We need to make necessary changes like dataset source path, models path etc., in a config file.

Finally, we can run the main file for adaptive threshold

python adaptive-threshold-main.py

Since the project was completed within a short period of time, lot of code optimization for faster inference is needed.

If you have any doubts, please inbox me.