The Labeled Ancestral Origin Faces (LAOFIW) is a dataset of 14,000 images divided into four equally sized classes: sub-Saharan Africa, East Asia, Indian subcontinent, Western Europe. The dataset was created to study the effects of bias in face recognition training datasets.
The authors used Bing Search API to query for origin search terms such as German, English, Polish for man, boy, woman, girl. "In total, 43 origin search terms were queried returning 20,000 images". The remaining 14,000 images, were manually divided into four broad ancestral origins: sub-Saharan Africa, Indian Subcontinent, Europe, and East Asia. These classes were selected on the basis of being visually distinct from each other. The URLs for these images are included in the dataset. It includes at least 108 images from Flickr.com.
The author re-released the list of ancestral images again, fixing several broken links.
The class labels only include 3 classes:
If you reference or use any data from the Exposing.ai project, cite our original research as follows:
@online{Exposing.ai, author = {Harvey, Adam. LaPlace, Jules.}, title = {Exposing.ai}, year = 2021, url = {https://exposing.ai}, urldate = {2021-01-01} }
If you reference or use any data from LAOFIW cite the author's work:
@inproceedings{Alvi2018TurningAB, author = "Alvi, Mohsan S. and Zisserman, Andrew and Nell{\aa}ker, Christoffer", title = "Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddings", booktitle = "ECCV Workshops", year = "2018" }