Service
Amazon Rekognition
API Action / Feature
Sync image-analysis actions (JSON 1.1, protocol RekognitionService):
DetectLabels — Image* → Labels[] (Name, Confidence, Instances, Parents, Aliases, Categories)DetectFaces — Image* → FaceDetails[] (BoundingBox, AgeRange, Emotions, Pose, Quality, etc.)DetectText — Image* → TextDetections[] (DetectedText, Type LINE/WORD, Confidence, Geometry)CompareFaces — SourceImage*, TargetImage* → FaceMatches[]/UnmatchedFaces[]DetectModerationLabels — Image* → ModerationLabels[]
Image is accepted as Bytes or S3Object, same "accept but don't decode" pattern floci already uses for Textract's Document input. The remaining ~70 operations (face-collection persistence: CreateCollection/IndexFaces/SearchFaces; custom-model training: Projects/Datasets/ProjectVersions; live video: StreamProcessor; async video jobs: Start*/Get* Celebrity/Content/Face/Label/Person/Segment/Text detection; Face Liveness sessions) are out of scope for this issue — each is a meaningfully bigger, stateful piece of work.
AWS Documentation
https://docs.aws.amazon.com/rekognition/latest/APIReference/Welcome.html
Why is this needed?
Rekognition has no coverage in floci today. Like Comprehend, it's a widely-used AWS AI service that shows up often in interviews and real applications (image moderation pipelines, content tagging, face comparison for identity flows), but currently can't be developed or tested locally against floci. The five sync Detect*/CompareFaces actions above are self-contained (no persisted state, no async job lifecycle, no video/streaming) and follow the exact same JSON 1.1 wire pattern already used for Comprehend, making this a well-scoped, precedent-following addition.
Are you willing to contribute a PR?
Service
Amazon Rekognition
API Action / Feature
Sync image-analysis actions (JSON 1.1, protocol
RekognitionService):DetectLabels— Image* → Labels[] (Name, Confidence, Instances, Parents, Aliases, Categories)DetectFaces— Image* → FaceDetails[] (BoundingBox, AgeRange, Emotions, Pose, Quality, etc.)DetectText— Image* → TextDetections[] (DetectedText, Type LINE/WORD, Confidence, Geometry)CompareFaces— SourceImage*, TargetImage* → FaceMatches[]/UnmatchedFaces[]DetectModerationLabels— Image* → ModerationLabels[]Imageis accepted asBytesorS3Object, same "accept but don't decode" pattern floci already uses for Textract'sDocumentinput. The remaining ~70 operations (face-collection persistence:CreateCollection/IndexFaces/SearchFaces; custom-model training:Projects/Datasets/ProjectVersions; live video:StreamProcessor; async video jobs:Start*/Get*Celebrity/Content/Face/Label/Person/Segment/Text detection; Face Liveness sessions) are out of scope for this issue — each is a meaningfully bigger, stateful piece of work.AWS Documentation
https://docs.aws.amazon.com/rekognition/latest/APIReference/Welcome.html
Why is this needed?
Rekognition has no coverage in floci today. Like Comprehend, it's a widely-used AWS AI service that shows up often in interviews and real applications (image moderation pipelines, content tagging, face comparison for identity flows), but currently can't be developed or tested locally against floci. The five sync Detect*/CompareFaces actions above are self-contained (no persisted state, no async job lifecycle, no video/streaming) and follow the exact same JSON 1.1 wire pattern already used for Comprehend, making this a well-scoped, precedent-following addition.
Are you willing to contribute a PR?