Posts Tagged ‘In the wild’

VGG & VGG2

VGG & VGG2:

These two face recognition datasets contain color face images of celebrities collected from the web. The images are available with large variation of poses and ages for both datasets.

VGG

VGG has no overlap with some other popular benchmarks such as LFW. Because the images are subject to copyright and VGG does not own the images, only URLs of the images are available by VGG. More information and links for download can be found on http://www.robots.ox.ac.uk/~vgg/data/vgg_face/. Each celebrity’s name is the name of a text file containing the image URLs and corresponding face detections.

Here is some information regarding VGG dataset:

  • Number of identities: 2622

If you use this dataset:

Please make sure to use the dataset for non-commercial research purposes only (Terms of Use). The detailed Terms of Use can be found on http://www.robots.ox.ac.uk/~vgg/data/vgg_face/licence.txt).

Please make sure to cite the paper:

O. M. Parkhi, A. Vedaldi, A. Zisserman, Deep Face Recognition. British Machine Vision Conference, 2015.

VGG2

VGG2 provides loosely cropped faces in separated files to download for training and testing. More information and links for download can be found on http://www.robots.ox.ac.uk/~vgg/data/vgg_face2/data_infor.html. You will need to create an account to be able to download the files.

Here is some information regarding VGG2 dataset:

  • Number of identities: 9131 (8631 identities for training, 500 identities for testing)

  • More than 3.3 million images in the wild

  • Almost 362 image samples per person

If you use this dataset:

Please make sure to pay attention to the License information for using the dataset for Commercial/Research purposes (Terms of Use) available on http://www.robots.ox.ac.uk/~vgg/data/vgg_face2/.

Please make sure to cite the paper:

Q. Cao, L. Shen, W. Xie, O. M. Parkhi, A. Zisserman, VGGFace2: A Dataset for Recognizing Face across Pose and Age. International Conference on Automatic Face and Gesture Recognition, 2018.

keywords: VisionImage, Face, Face Verification, In the Wild

LFW: Labeled Faces in the Wild

LFW: Labeled Faces in the Wild:

This dataset contains labeled face images collected from the web with names of the people in the images as the labels. Some of these people have two or more number of images in the dataset. This dataset is designed for studying the problem of unconstrained face recognition and face verification. The original LFW dataset is available for download along with 3 sets of aligned images (funneled images, LFW-a, deep funneled).

Here is some information regarding this dataset:

  • Number of images in the dataset: 13,000 images (10-fold cross validation is recommended and training and test splits can be downloaded from the dataset page)

  • Number of identities: 5749

  • Image resolution: 250×250

More details and links for download can be found on the dataset page http://vis-www.cs.umass.edu/lfw/.

If you use the any of these versions of the LFW image dataset:

Please make sure to cite the paper:

G. B. Huang, M. Ramesh, T. Berg, E. Learned-Miller, Labeled Faces in the Wild: A Database for Studying Face Recognition in Unconstrained Environments. University of Massachusetts, Amherst, Technical Report 07-49, October 2007.

If you use the LFW imaged aligned by deep funneling:

Please make sure to cite the paper:

G. B. Huang, M. Matter, H. Lee, E. Learned-Miller, Learning to Align from Scratch. Advances in Neural Information Processing Systems (NIPS), 2012.

If you use the LFW imaged aligned by funneling:

Please make sure to cite the paper:

G. B. Huang, V. Jain, E. Learned-Miller, Unsupervised Joint Alignment of Complex Images. International Conference on Computer Vision (ICCV), 2007.

keywords: Vision, Image, Face, Object Detection, Segmentation, In the Wild