feat(model): add ocr output (#94)

Because

- support OCR task

This commit

- add OCR task output format

vdp/model/v1alpha/common.proto

@@ -0,0 +1,18 @@
+syntax = "proto3";
+
+package vdp.model.v1alpha;
+
+// Google api
+import "google/api/field_behavior.proto";
+
+// BoundingBox represents the bounding box data structure
+message BoundingBox {
+  // Bounding box top y-axis value
+  float top = 1 [ (google.api.field_behavior) = OUTPUT_ONLY ];
+  // Bounding box left x-axis value
+  float left = 2 [ (google.api.field_behavior) = OUTPUT_ONLY ];
+  // Bounding box width value
+  float width = 3 [ (google.api.field_behavior) = OUTPUT_ONLY ];
+  // Bounding box height value
+  float height = 4 [ (google.api.field_behavior) = OUTPUT_ONLY ];
+}

vdp/model/v1alpha/detection_output.proto

@@ -5,17 +5,7 @@ package vdp.model.v1alpha;
 // Google api
 import "google/api/field_behavior.proto";
 
-// BoundingBox represents the bounding box data structure
-message BoundingBox {
-  // Bounding box top y-axis value
-  float top = 1 [ (google.api.field_behavior) = OUTPUT_ONLY ];
-  // Bounding box left x-axis value
-  float left = 2 [ (google.api.field_behavior) = OUTPUT_ONLY ];
-  // Bounding box width value
-  float width = 3 [ (google.api.field_behavior) = OUTPUT_ONLY ];
-  // Bounding box height value
-  float height = 4 [ (google.api.field_behavior) = OUTPUT_ONLY ];
-}
+import "vdp/model/v1alpha/common.proto";
 
 // BoundingBoxObject represents a predicted bounding box object
 message BoundingBoxObject {

vdp/model/v1alpha/model.proto

@@ -16,6 +16,7 @@ import "vdp/model/v1alpha/model_definition.proto";
 import "vdp/model/v1alpha/classification_output.proto";
 import "vdp/model/v1alpha/detection_output.proto";
 import "vdp/model/v1alpha/keypoint_output.proto";
+import "vdp/model/v1alpha/ocr_output.proto";
 import "vdp/model/v1alpha/unspecified_task_output.proto";
 
 // Model represents a model
@@ -495,8 +496,10 @@ message ModelInstanceOutput {
     DetectionOutput detection = 2 [ (google.api.field_behavior) = OUTPUT_ONLY ];
     // The keypoint output
     KeypointOutput keypoint = 3 [ (google.api.field_behavior) = OUTPUT_ONLY ];
+    // The ocr output
+    OcrOutput ocr = 4 [ (google.api.field_behavior) = OUTPUT_ONLY ];
     // The unspecified task output
-    UnspecifiedTaskOutput unspecified = 4
+    UnspecifiedTaskOutput unspecified = 5
         [ (google.api.field_behavior) = OUTPUT_ONLY ];
   }
 }

vdp/model/v1alpha/ocr_output.proto

@@ -0,0 +1,19 @@
+syntax = "proto3";
+
+package vdp.model.v1alpha;
+
+import "vdp/model/v1alpha/common.proto";
+
+// Google api
+import "google/api/field_behavior.proto";
+
+// OcrOutput represents the output of ocr task
+message OcrOutput {
+  // A list of bounding box text
+  repeated BoundingBox bounding_boxes = 1
+      [ (google.api.field_behavior) = OUTPUT_ONLY ];
+  // A list of ocr texts
+  repeated string texts = 2
+      [ (google.api.field_behavior) = OUTPUT_ONLY ];
+}
+