emmc_python/Aida_data/sfe/doc/frame_data.proto

syntax = "proto3";
package innovatrics.embedded.frame_data;
import "google/protobuf/timestamp.proto";

message FrameData {
    // timestamp of the frame the message was generated for
    google.protobuf.Timestamp timestamp = 1;
    // incremental frame number
    uint64 frame_number = 2;
    // full frame image
    Image frame_image = 3;
    // vector of detected faces
    repeated FaceData face_data = 4;
    // vector of lost objects
    repeated LostObjectData lost_object_data = 5;
    // client_id (same as MQTT client ID) of the source camera/device which produced the message
    string client_id = 6;
}

message FaceData {
    // face detection data
    FaceDetectionData face_detection_data = 1;
    // face tracking data
    TrackingData face_tracking_data = 2;
    // face landmarks data
    repeated FaceLandmarkData landmarks_data = 3;
    // face mask status data
    optional FaceMaskData face_mask_data = 4;
    // cropping data
    optional CroppingData cropping_data = 5;
    // liveness detection data 
    repeated LivenessData liveness_data = 6;
    // face template extraction
    optional TemplateData template_data = 7;
    // face identification
    repeated IdentificationData identification_data = 8;
}

message FaceDetectionData {
    // bounding box of detected face
    BoundingBox bounding_box = 1;
    // normalized value detection confidence of detected face - range <0,1>
    float detection_confidence = 2;
    // raw value of detection confidence of detected face - range <0,1>
    float raw_detection_confidence = 3;
}

message TrackingData {
    // tracking ID (ByteTrack) assigned to detected object
    uint32 tracking_id = 1;
    // tracking UUID assigned to detected object
    string tracking_uuid = 2;
    // tracking state (ByteTrack) of detected object
    TrackingState tracking_state = 3;
}

message FaceLandmarkData {
    // type of the keypoint detected in the face
    FaceKeypointType keypoint_type = 1;
    // confidence of the keypoint detected in the face - range <0,1>
    float confidence = 2;
    // X coordinate of the keypoint relative to full frame - range <0,1>
    float x = 3;
    // Y coordinate of the keypoint relative to full frame - range <0,1>
    float y = 4;
}

message FaceMaskData {
    // face mask detection confidence - range <0,1>
    float confidence = 1;
}

message CroppingData {
    // extension of the crop image specified in settings
    uint32 crop_extension = 1;
    // bounding box of the crop
    optional BoundingBox crop_box = 2;
    // crop image of detected object
    optional Image crop_image = 3;
}

message LivenessData {
    // type of liveness detection algorithm
    LivenessType liveness_type = 1;
    // liveness detection score
    float score = 2;
}

message TemplateData {
    bytes template = 1;
}

message IdentificationData {
    // user id
    uint32 id = 1;
    // user name
    string name = 2;
    // match score
    float score = 3;
}

message Image {
    // width of the image - absolute value
    uint32 width = 1;
    // height of the image - absolute value
    uint32 height = 2;
    // image data in specified format
    optional bytes data = 3;
    // format of the image data specified in settings
    ImageFormat image_format = 4;
}

// image format enumeration
enum ImageFormat {
    // default
    RawBgr = 0;
    // Jpeg
    Jpeg = 1;
    // Png
    Png = 2;
}

message BoundingBox {
    // X coordinate of left upper corner of bounding box relative to full frame - range <0,1>
    float x = 1;
    // Y coordinate of left upper corner of bounding box relative to full frame - range <0,1>
    float y = 2;
    // width of bounding box relative to width of full frame - range <0,1>
    float width = 3;
    // height of bounding box relative to height of full frame - range <0,1>
    float height = 4;
}

// tracking state enumeration (defined by ByteTrack)
enum TrackingState {
    New = 0;
    Tracked = 1;
    Lost = 2;
    Removed = 3;
}

// face keypoint type enumeration
enum FaceKeypointType {
    RightEyeOuterCorner = 0;
    RightEyeCentre = 1;
    RightEyeInnerCorner = 2;
    LeftEyeInnerCorner = 3;
    LeftEyeCentre = 4;
    LeftEyeOuterCorner = 5;
    NoseRoot = 6;
    NoseRightBottom = 7;
    NoseTip = 8;
    NoseLeftBottom = 9;
    NoseBottom = 10;
    MouthRightCorner = 11;
    MouthCenter = 12;
    MouthLeftCorner = 13;
    MouthUpperEdge = 14;
    MouthLowerEdge = 15;
    RightEyebrowOuterEnd = 16;
    RightEyebrowInnerEnd = 17;
    LeftEyebrowInnerEnd = 18;
    LeftEyebrowOuterEnd = 19;
    RightEdge = 20;
    ChinTip = 21;
    LeftEdge = 22;
}

// liveness detection algorithm enumeration
enum LivenessType {
    // default
    Distant = 0;
}

message LostObjectData {
    // tracking UUID of the lost object
    string tracking_uuid = 1;
    // timestamp of the frame where the object appeared first time (New in ByteTrack)
    google.protobuf.Timestamp first_time_appeared = 2;
    // timestamp of the frame where ByteTrack change the tracking status of the object to Removed
    google.protobuf.Timestamp last_time_appeared = 3;
    // crop of best appearance of the object in tracklet - NOT SUPPORTED YET
    optional Image best_crop_image = 4;
}