New face detection algorithm, processing arbitrary face orientations.
Face thumbnail in responses, cropped and properly aligned.
Rotation angle in response.
Face attributes: age(child/adult), emotion(neutral/smile).
Face ROI in response for face matching for unambiguous results interpretation.
Face thumbnail in responses.
Updated algorithms for liveness classification.
Rejecting faces with occlusions (medicine mask, sunglasses, hat, hands).
Rejecting art face masks.
Rejecting faces with closed eyes.
Updated model for classification.
WARNING: app validation logic returns 200 instead of 400 http codes. Previously, if you supply bad images, or wrong params, you will get response with 400 http code. Starting from this version, you will always get 200 result and should check code property inside response json-body to get operation result. All possible codes are listed here
WARNING: all project related libraries changed their names and artifacts names to FaceSDK. Now project name is consistent across all platforms.
WARNING: /detect now by default return only central face.
face compare operation renamed to face match operation. This leads to changing model names in clients. /compare endpoint renamed to /match endpoint. Old models and routes are still present via type aliases for backward compatibility.
Migrate to python 3.8.
[docker] by default do not spam access log to stdout.
[docker] send all apps logs to stdout instead of stderr.
[docker] logs format changed.
[linux,win] by default store app logs into daily rotating "logs/app/facesdk-app.log" file.
HTTPS certs custom path options for all platforms (previously was only in docker).
Process logs now support try_id params(beside custom id). Now operations inputs and results saved under next id
invocation_id = uuid.uuid4().hex if not id else id + '_' + try_id.
C# client published.
Logging format unified across all platforms.
Added ability to store logs into files with daily rotation.
JSON log formatting (useful for log collectors).