Skip to content

Tater v1.1.22

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 01 Sep 19:05
· 1 commit to main since this release
v1.1.22
23333e9

Tater v1.1.22

Tater v1.1.22 adds AdaFace as a selectable Face ID model while preserving
FaceNet compatibility, safe rollback, and centralized Spud Hub processing.

What's Changed

AdaFace Face Recognition

  • Adds the official AdaFace IR-50 WebFace4M model as an experimental Face ID
    option alongside FaceNet512.
  • Adds a recognition-model selector and migration progress to Settings ›
    Models › Face ID.
  • Pins the AdaFace checkpoint revision and validates its required runtime
    dependencies across macOS, CPU Docker, NVIDIA Docker, and private setup.
  • Uses RetinaFace alignment and normalized 512-dimensional, model-tagged
    embeddings with a conservative AdaFace matching threshold.
  • Adds a labeled-image bakeoff utility for comparing genuine and impostor
    distances between FaceNet and AdaFace using real camera images.

Safe Model Switching

  • Re-embeds saved face crops in the background before activating a newly
    selected model.
  • Keeps separate FaceNet and AdaFace embedding profiles and never compares
    vectors produced by different models.
  • Preserves the previous model's embeddings for immediate rollback and only
    generates embeddings that are missing when switching again.
  • Leaves the current model active and reports an error if every linked person
    cannot receive a usable embedding from the requested model.

SpudLink Face ID Synchronization

  • Lets Spudlets detect the Spud Hub's active Face ID model from tagged
    embedding responses.
  • Automatically re-embeds saved crops for linked people through the Hub when
    the Hub changes models, then resumes recognition with compatible vectors.
  • Keeps Face ID model execution and downloads on the Hub; connected Spudlets
    store only the returned embeddings and can retain FaceNet as a local
    fallback.

Updating

  • macOS users already running v1.0.1 or later can install v1.1.22 through
    Tater's normal updater after its signed macOS package is published.
  • macOS users still running v100 or earlier must perform the one-time manual
    app replacement described with v1.0.1 because those builds treat the new
    semantic version as older than 100.
  • Docker users can pull v1.1.22 or latest for the CPU image and
    v1.1.22-nvidia or nvidia for the NVIDIA image after the release tag is
    published.
Assets 5
Loading