Publish the official YuNet 2026 export with verified hashes, tensor contracts and padding requirements. Retain the earlier static detector for reproducing previous benchmarks.
Add verified TinyFaceMatch, AdaFace FP16 and YuNet exports
Publish static batch-one exports with float32 public I/O, SHA-256 manifest, benchmark notes and upstream model terms. AdaFace is the optional larger model for Ray Local; TinyFaceMatch is the default.
