10,000+ videos of face presentation attacks using 12+ premium resin masks, custom-manufactured to replicate specific individuals' facial features. This dataset, created by AxonData, is designed for iBeta Level 3 certification testing and was last updated on 2026-04-24.
Use Cases
- Training face presentation attack detection (PAD) models based on videos of high-precision resin mask attacks.
- Benchmarking anti-spoofing algorithm robustness against the most advanced attack type in the PAD taxonomy.
- Conducting certification testing for iBeta Level 3 compliance based on the specified attack videos.
Strengths
- Contains over 10,000 attack videos, providing substantial volume for model training and testing.
- Focuses on 12+ premium resin masks, which are described as the most advanced and realistic attack type.
- Masks are custom-manufactured to replicate specific individuals, creating a high-precision and challenging detection scenario.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- AxonData
- Collection Method
- Likely involves recording video presentations of custom-manufactured resin masks.
- Time Range
- null
- Freshness
- Last updated 2026-04-24 08:35:53; freshness should be verified.
- Geography
- null