Awesome

vgg-faces-utils

Script to download and annotate images from VGG Faces dataset

Usage

Download resources

Download VGG Face Dataset from it's home page:

http://www.robots.ox.ac.uk/~vgg/data/vgg_face/

Or:

https://mega.nz/#!zeQSwC7D!aOBgSYkvldL6k91VdwKIx_EKS1J3_MQduPEU3QQ8g-o

You would get a file vgg_face_dataset.tar.gz that contains a vgg_face_dataset folder.

Setup tools

git clone https://github.com/ndaidong/vgg-faces-utils.git

# extract compressed file into repo folder
tar -zxvf vgg_face_dataset.tar.gz -C vgg-faces-utils


cd vgg-faces-utils

# create virtual environment
python3 -m venv venv
source venv/bin/activate

# install packages
(venv) pip install -r requirements.txt

# start processing
(venv) ./script.py -d vgg_face_dataset/files -o output

The process may take several days depending on the connectivity.

It would try to download about 2.6 milions images from many resources. However, we just expect to get about 1 milion. Almost of links no longer available for right now.

Update annotations with labelImg

To check and modify the labels after downloading finishes.

(venv) labelImg

Specify image dir and label dir to get:

License

The MIT License (MIT)