Research

Open protocols and reproducible face-analysis checks

iLook research separates repeatability and photo sensitivity from unsupported claims about beauty, identity, health, or personal worth.

Face Analysis Stability Benchmark 2026

The preregistered protocol defines a 100-image licensed synthetic adult evaluation set, five repeated baseline runs, and controlled photo-condition tests. Results are pending and no benchmark statistics are claimed yet.

Read the benchmark protocol

Research principles

  • Publish methods before results.
  • Release only aggregate, non-identifying measurements.
  • Version the model and scoring rubric used for every run.
  • Describe limitations and negative results.
  • Never use private user photos without separate, explicit research consent.