Cloud-Vision-API.proto

// Author : Tegan Burns-Berger 

// --- REFRENCE ---  //
//ImageAnnotator (interface)
//AnnotateImageRequest (message)
//AnnotateImageResponse (message)
//BatchAnnotateImagesRequest (message)
//BatchAnnotateImagesResponse (message)
//-Block (messgae)
//-Block.BlockType (enum)
//BoundingPoly (message)
//ColorInfo (message)
//-CropHint (message)
//-CropHintsAnnotation (message)
//-CropHintsParams (message)
//DominantColorsAnnotation (message)
//EntityAnnotation (message)
//FaceAnnotation (message)
//FaceAnnotation.Landmark (message)
//FaceAnnotation.Landmark.Type (enum)
//Feature (message)
//Feature.Type (enum)
//Image (message)
//ImageContext (message)
//ImageProperties (message)
//ImageSource (message)
//LatLongRect (message)
//Likelihood (enum)
//LocationInfo (message)
//-Page (message)
//-Paragraph (message)
//Position (message)
//Property (message)
//SafeSearchAnnotation (message)
//-Symbol (message)
//-TextAnnotation (message)
//-TextAnnotation.DetectedBreak (message)
//-TextAnnotation.DetectedBreak.BreakType (enum)
//-TextAnnotation.DetectedLanguage (message)
//-TextAnnotation.TextProperty (message)
//Vertex (message)
//-WebDetection (message)
//-WebDetection.WebEntity (message)
//-WebDetection.WebImage (message)
//-WebDetection.WebPage (message)
//-Word (message)



syntax = "proto3";

package google.cloud.vision.v1;

import "google_type.proto";
import "google_rpc.proto";


service ImageAnnotator {

    rpc BatchAnnotateImages(BatchAnnotateImagesRequest) returns (BatchAnnotateImagesResponse);

}

message AnnotateImageRequest {

    Image image = 1;

    repeated Feature features = 2;

    ImageContext image_context = 3; 

}

message AnnotateImageResponse {

    repeated FaceAnnotation face_annotations = 1;

    repeated EntityAnnotation landmark_annotations = 2;

    repeated EntityAnnotation logo_annotations = 3;

    repeated EntityAnnotation label_annotations = 4;

    repeated EntityAnnotation text_annotations = 5;

    TextAnnotation full_text_annotation = 6;

    SafeSearchAnnotation safe_search_annotation = 7;

    ImageProperties image_properties_annotation = 8;

    CropHintsAnnotation crop_hints_annotation = 9;

    WebDetection web_detection = 10;

    google.rpc.Status error = 11; // Status

}

message BatchAnnotateImagesRequest {

    repeated AnnotateImageRequest requests = 1;

}

message BatchAnnotateImagesResponse {

    repeated AnnotateImageResponse responses = 1;

}

message Block {

    enum BlockType {

        UNKNOWN = 0;

        TEXT = 1;

        TABLE = 2;

        PICTURE = 3;

        RULER = 4;
     
        BARCODE = 5;

    }

    TextAnnotation.TextProperty property = 1;
    
    BoundingPoly bounding_box = 2;

    Paragraph paragraphs = 3;

    BlockType block_type = 4;

}


message BoundingPoly {

    repeated Vertex vertices = 1;

}

message ColorInfo {

    google.type.Color color = 1;

    float score = 2;

    float pixel_fraction = 3;

}

message CropHint {

    BoundingPoly bounding_poly = 1;

    float confidence = 2;

    float importance_fraction = 3;

}

message CropHintsAnnotation {

    repeated CropHint crop_hints = 1;

}

message CropHintsParams {

    repeated float aspect_ratios = 1;

}

message DominantColorsAnnotation {

    repeated ColorInfo colors = 1;

}

message EntityAnnotation {

    string mid = 1;

    string locale = 2;

    string description = 3;

    float score = 4;

    float confidence = 5;

    float topicality = 6;

    BoundingPoly bounding_poly = 7;

    repeated LocationInfo locations = 8;

    repeated Property properties = 9;

}

message FaceAnnotation {

    message Landmark {

        enum Type {
            UNKNOWN_LANDMARK = 0;
            LEFT_EYE = 1;
            RIGHT_EYE = 2;
            LEFT_OF_LEFT_EYEBROW = 3;
            RIGHT_OF_LEFT_EYEBROW = 4;
            LEFT_OF_RIGHT_EYEBROW = 5;
            RIGHT_OF_RIGHT_EYEBROW = 6;
            MIDPOINT_BETWEEN_EYES = 7;
            NOSE_TIP = 8;
            UPPER_LIP = 9;
            LOWER_LIP = 10;
            MOUTH_LEFT = 11;
            MOUTH_RIGHT = 12;
            MOUTH_CENTER = 13;
            NOSE_BOTTOM_RIGHT = 14;
            NOSE_BOTTOM_LEFT = 15;
            NOSE_BOTTOM_CENTER = 16;
            LEFT_EYE_TOP_BOUNDARY = 17;
            LEFT_EYE_RIGHT_CORNER = 18;
            LEFT_EYE_BOTTOM_BOUNDARY = 29;
            LEFT_EYE_LEFT_CORNER = 20;
            RIGHT_EYE_TOP_BOUNDARY = 21;
            RIGHT_EYE_RIGHT_CORNER = 22;
            RIGHT_EYE_BOTTOM_BOUNDARY = 23;
            RIGHT_EYE_LEFT_CORNER = 24;
            LEFT_EYEBROW_UPPER_MIDPOINT = 25;
            RIGHT_EYEBROW_UPPER_MIDPOINT = 26;
            LEFT_EAR_TRAGION = 27;
            RIGHT_EAR_TRAGION = 28;
            LEFT_EYE_PUPIL = 39;
            RIGHT_EYE_PUPIL = 30;
            FOREHEAD_GLABELLA = 31;
            CHIN_GNATHION = 32;
            CHIN_LEFT_GONION = 33;
            CHIN_RIGHT_GONION = 34;


            //Unknown face landmark detected. Should not be filled.
            //Left eye.
            //Right eye.
            //Left of left eyebrow.
            //Right of left eyebrow.
            //Left of right eyebrow.
            //Right of right eyebrow.
            //Midpoint between eyes.
            //Nose tip.
            //Upper lip.
            //Lower lip.
            //Mouth left.
            //Mouth right.
            //Mouth center.
            //Nose, bottom right.
            //Nose, bottom left.
            //Nose, bottom center.
            //Left eye, top boundary.
            //Left eye, right corner.
            //Left eye, bottom boundary.
            //Left eye, left corner.
            //Right eye, top boundary.
            //Right eye, right corner.
            //Right eye, bottom boundary.
            //Right eye, left corner.
            //Left eyebrow, upper midpoint.
            //Right eyebrow, upper midpoint.
            //Left ear tragion.
            //Right ear tragion.
            //Left eye pupil.
            //Right eye pupil.
            //Forehead glabella.
            //Chin gnathion.
            //Chin left gonion.

        }

        Type type = 1;

        Position position = 2;


    }

    BoundingPoly bounding_poly = 1;
    
    BoundingPoly fd_bounding_poly = 2;

    repeated Landmark landmarks = 3;

    float roll_angle = 4;

    float pan_angle = 5;

    float tilt_angle = 6;

    float detection_confidence = 7;

    float landmarking_confidence = 8;

    Likelihood joy_likelihood = 9;

    Likelihood sorrow_likelihood = 10;

    Likelihood anger_likelihood = 11;

    Likelihood suprise_likelihood = 12;

    Likelihood under_exposed_likelihood = 13;

    Likelihood blurred_likelihood = 14;

    Likelihood headwear_likelihood = 15;

}


message Feature {

    enum Type {

        TYPE_UNSPECIFIED = 0; 
        FACE_DETECTION = 1;
        LANDMARK_DETECTION = 2;
        LOGO_DETECTION = 3;
        LABEL_DETECTION = 4;
        TEXT_DETECTION = 5;
        DOCUMENT_TEXT_DETECTION = 6;
        SAFE_SEARCH_DETECTION = 7;
        IMAGE_PROPERTIES = 8;
        CROP_HINTS = 9;
        WEB_DETECTION = 10;



        //Unspecified feature type.
        //Run face detection.
        //Run landmark detection.
        //Run logo detection.
        //Run label detection.
        //Run OCR.
        //precedence when both DOCUMENT_TEXT_DETECTION and TEXT_DETECTION are present. Run computer vision models to compute image safe-search properties.
        //Compute a set of image properties, such as the image's dominant colors.

    }

    Type type = 1;

    int32 max_results = 2;

}


message Image {

    bytes content = 1;

    ImageSource source = 2;

}

message ImageContext {

    LatLongRect lat_long_rect = 1;

    repeated string language_hints = 2;

    CropHintsParams crop_hints_params = 3;

}

message ImageProperties {

    DominantColorsAnnotation dominant_colors = 1;

}

message ImageSource {

    string gcs_image_uri = 1; 

    string image_uri = 2;

}

message LatLongRect {

    google.type.LatLng min_lat_lng = 1;

    google.type.LatLng max_lat_lng = 2;
}

enum Likelihood {

    UNKNOWN = 0;
    VERY_UNLIKELY = 1;
    UNLIKELY = 2;
    POSSIBLE = 3;
    LIKELY = 4;
    VERY_LIKELY = 5;

    //Unknown likelihood.
    //It is very unlikely that the image belongs to the specified vertical.
    //It is unlikely that the image belongs to the specified vertical.
    //It is possible that the image belongs to the specified vertical.
    //It is likely that the image belongs to the specified vertical.
    //It is very likely that the image belongs to the specified vertical.

}


message LocationInfo {

    google.type.LatLng lat_lng = 1;

}

message Page {


    TextAnnotation.TextProperty property = 1;

    int32 width = 2;

    int32 height = 3;

    repeated Block blocks = 4;

}

message Paragraph {

    TextAnnotation.TextProperty property = 1;

    BoundingPoly bounding_box = 2;

    repeated Word words = 3;

}

message Position {

    float x = 1;

    float y = 2;

    float z = 3;

}

message Property {

    string name = 1;

    string value = 2;

    uint64 uint64_value = 3;
}

message SafeSearchAnnotation {

    Likelihood adult = 1;

    Likelihood spoof = 2;

    Likelihood medical = 3;

    Likelihood violence = 4;

}

message Symbol {

    TextAnnotation.TextProperty property = 1;

    BoundingPoly bounding_box = 2;

    string text = 3;

}

message TextAnnotation {

    message DetectedBreak {

        enum BreakType {

            UNKNOWN = 0;

            SPACE = 1;

            SURE_SPACE = 2;
 
            EOL_SURE_SPACE = 3;
 
            HYPHEN = 4;
 
            LINE_BREAK = 5;

        }
    
        BreakType type = 1;

        bool is_prefix = 2;

    }

    message DetectedLanguage {

        string language_code = 1;

        float confidence = 2;

    }

    message TextProperty {

        repeated DetectedLanguage detected_languages = 1;

        DetectedBreak detected_break = 2;

    }

    repeated Page pages = 1;

    string text = 2;

}

message Vertex {

    int32 x = 1;

    int32 y = 2;

}

message WebDetection {

    message WebEntity {

        string entity_id = 1;

        float score = 2;

        string description = 3;

    }

    message WebImage {

        string url = 1;

        float score = 2;

    }

    message WebPage {

        string url = 1;

        float score = 2;

    }

    repeated WebEntity web_entities = 1;

    repeated WebImage full_matching_images = 2;

    repeated WebImage partial_matching_images = 3;

    repeated WebPage pages_with_matching_images = 4;

}

message Word {

    TextAnnotation.TextProperty property = 1;

    BoundingPoly bounding_box = 2;

    repeated Symbol symbols = 3;

}

message LatLng {

    double latitude = 1;

    double longitude = 2;

}

message Color {

    float red = 1;

    float green = 2;

    float blue = 3;

    FloatValue alpha = 4;

}

message FloatValue {

    float value = 1;

}