Research-driven datasets for face anti-spoofing, liveness detection, face recognition, and voice biometrics

axonlab.ai · sales@axonlabs.pro

Axon Labs builds training and evaluation datasets for biometric AI systems. Our data powers face recognition, liveness detection , iBeta certification preparation, KYC / eKYC verification, and voice biometrics.

We are used by 50+ clients — fintech companies, eKYC platforms, biometric SDK vendors, and security teams — and our data has contributed to 21% of iBeta 2025 certified solutions.

• 180,000+ videos across all datasets
• 30+ ready-to-license datasets
• 1,000+ unique participants, diverse demographics and devices
• All data collected with signed consent and GDPR-compliant

Every dataset page (repository) in this organization answers the six questions our clients ask most:

1. Number of unique IDs — how many distinct subjects
2. Number of videos / images — total volume
3. Variability — lighting, environments, backgrounds, accessories, attributes
4. Real-to-spoof pairing — whether genuine face videos are paired with matching attack videos
5. Demographics — gender, age, ethnicity distribution
6. Devices — how many and which specific device models were used for capture

Training and certification-grade datasets for ISO/IEC 30107-3 compliant systems and iBeta Level 1 / Level 2 / Level 3 preparation.

• ibeta-level-1-face-anti-spoofing-dataset — 30,000+ PAD attack videos (paper, cutout, replay) from 85+ participants, ISO/IEC 30107-3 compliant
• partial-paper-mask-face-anti-spoofing-dataset — 3,000 videos, 50 participants, dual-device capture
• cardboard-mask-face-anti-spoofing-dataset — cardboard mask attacks with real accessories (wigs, glasses, hats), 3,000 videos, 50 participants, multi-device capture

Additional PAD datasets we can share on request: silicone masks, latex masks, cloth masks, 3D resin masks, photo print attacks, replay display attacks, high-fidelity mask variants.

• Selfie_and_Official_ID_Photo_Dataset — 6,000+ people, 70,000+ images: 10–15 photos per ID (selfies + 2 official ID photos). For face recognition, KYC verification, identity matching, biometric model training. Ages 18–65, balanced demographics.
• human-faces-dataset-multiple-images — 1,000+ people, 10,000+ files: 8 photos per person + 2 videos each
• age-estimation-minors-face-dataset — 10,000+ consented selfies of minors and young adults (10–30 years) with verified per-year age labels. Multi-ethnic, phone-captured. For under-18 age gating, age verification, age estimation models.

Browse all repositories in the Repositories tab.

• iBeta PAD certification (Level 1, 2, 3) preparation and auditing
• eKYC and onboarding — selfie vs. ID matching, liveness gating, age verification
• Face recognition training for fintech, banking, and telecom
• Deepfake and presentation attack detection (PAD) research
• Under-18 age gating for regulated platforms
• Voice biometric authentication in contact centers

• All participants sign informed consent before capture
• GDPR-compliant processing and storage
• Commercial licensing — no web-scraped data
• Attack datasets collected in controlled conditions with participant release

• Sales & licensing: sales@axonlabs.pro
• Website: axonlab.ai