yohannesb/Ethiopian-Face-Dataset

The Ethiopian Face Dataset is the first publicly available unconstrained face image dataset collected in Ethiopia. It is designed to support research in face recognition, computer vision, fairness analysis, and AI applications focusing on underrepresented populations. # Ethiopian Face Dataset The Ethiopian Face Dataset is the first publicly available unconstrained face image dataset collected in Ethiopia. It is designed to support research in face recognition, computer vision, fairness analysis, and AI applications focusing on underrepresented populations. ## Overview This dataset consists of 450 unconstrained face images representing 45 Ethiopian individuals, each with 10 images captured under real-world conditions. The dataset reflects natural variations commonly encountered in practical face recognition scenarios, including: * Illumination changes * Diverse backgrounds * Pose variations * Facial expressions * Occlusions * Makeup * Aging (for some subjects) ## Data Collection Process The gender distribution includes 25 male and 20 female subjects. The dataset was constructed by extracting frames from a variety of Amharic movies and TV series, ensuring rich visual diversity. Sources include: * **TV Series:** Sewlesew, Gemena I & II, Gorebetamochu * **Standalone Films:** Wodegedelew, Kebrone, Zeraf, and others TV series were particularly valuable due to their wide range of scenes, lighting conditions, and facial appearances for the same character, making them ideal for building an unconstrained face recognition benchmark. ## Key Features * First unconstrained Ethiopian face dataset * 450 images, 45 subjects (10 images per subject) * Real-world variability (illumination, pose, background, occlusion, expression) * Useful for face recognition, domain adaptation, and fairness studies * Captures underrepresented facial morphology often missing in global datasets ## Intended Use This dataset is suitable for: * Face recognition algorithm evaluation * Deep learning model training and testing * Fairness and bias assessment in face analysis systems * Research on underrepresented demographic groups ## Citation If you use this dataset in your research, please cite the repository and the follwing paper. **BibTeX** ```bibtex @INPROC …