FDDB icon indicating copy to clipboard operation
FDDB copied to clipboard

Face Detection Data Set and Benchmark (FDDB) in Darknet

FDDB

This repository contains scripts automating conversion of the FDDB dataset to a format required by the Darknet framework.

Usage

  1. Clone and navigate to the repository

  2. Use the following links in order to download a complete FDDB dataset.

     wget http://vis-www.cs.umass.edu/fddb/FDDB-folds.tgz
     wget http://vis-www.cs.umass.edu/fddb/README.txt
     wget http://tamaraberg.com/faceDataset/originalPics.tar.gz
    
  3. Create a new folder and call it images. Unpack the supplied images inside of it.

     tar -xzvf FDDB-folds.tgz
     mdkir images && tar -xzvf originalPics.tar.gz -C images
    
  4. Prepare data.

     find FDDB-folds -type f -regex ".*[0-9]-.*txt" -exec cat {} >> FDDB-folds/FDDB-annotations.txt \;
     find FDDB-folds -type f -regex ".*[0-9]+.txt" -exec cat {} >> FDDB-folds/FDDB-paths.txt \;
    
  5. Install the required Python libraries.

     pip install Pillow
    
  6. Generate all the files required by the Darknet.

     python ./scripts/cs_fddb_convert_to_darknet.py
    

License

This is free and unencumbered software released into the public domain. For more information, please refer to the LICENSE file.