Off-the-shelf facial and biometric datasets for face recognition, KYC onboarding, age estimation and demographic bias testing.
Black people face recognition dataset - a large human faces dataset of more than 10 million images for face recognition models. Listed under facial recognition and biometrics datasets, and suitable for face recognition and verification, facial attribute analysis and demographic bias evaluation.
Facial images dataset of Asian participants - 44,240 images. Listed under facial recognition and biometrics datasets, and suitable for face detection and recognition, facial attribute analysis and demographic bias evaluation in biometric models.
A historical face recognition dataset built from archive photos taken between 6 months and more than 10 years before the current images - real photographs, not synthetic aging. It captures natural aging progression, including wrinkles, hairstyle changes and weight fluctuations.
Historical facial images and selfie dataset - 60,000 images covering African, Indian and South East Asian participants. Listed under facial recognition and biometrics datasets, and suitable for age-invariant face recognition, identity matching over time, and KYC or identity verification model training.
Human faces dataset containing both images and videos of participants of Asian and European ethnicity. Listed under facial recognition and biometrics datasets, and suitable for face detection, face recognition and verification, and multimodal biometric model training.
Japanese facial images, with and without makeup, across various poses and angles, including sports movements and activities such as walking, running and standing. Japanese voices with emotions are also included.
Kids age progression dataset - facial images of 1,998 child participants of mixed ethnicity. Listed under facial recognition and biometrics datasets, and suitable for age estimation, age-invariant face recognition and age progression model training.
Smartphone-captured facial images of 10,000 Indian participants with a 1:1 male-to-female ratio. Each person is photographed around 100 times at 30 cm, 50 cm and 80 cm, across three angles (15-45 degrees), indoors and outdoors, under varied lighting and with different props. Built for face recognition, liveness detection and mobile onboarding model training.
A selfie and official ID photo dataset covering more than 6,000 people and over 70,000 images, with 10 to 15 photos per identity (selfies plus two official ID photos). Well suited to face recognition, KYC verification, identity matching and biometric training, with participants aged 18 to 65 and balanced demographics.
Selfie and video dataset - more than 1,000 people and over 10,000 files, with 8 photos and 2 videos per person. Listed under facial recognition and biometrics datasets, and suitable for face recognition and verification, liveness detection, and KYC or identity onboarding model training.
Children are missing from most deepfake prevention and age detection datasets. This underrepresentation leads to biased models, weaker age-restricted access controls and increased vulnerability to deepfake attacks. The Synthetic Children Faces Dataset fills that gap: it expands demographic coverage while avoiding any real-person data.
