Change Landmark's landmark_type, landmark_coordinate, and decoding_type to string

A user can use the string while understanding purpose of
landmark_type, landmark_coordinate, and decoding_type metadata.
The string values are parsed and converted to enumeration type
inference_landmark_type_e, inference_landmark_coordinate_type_e,
and inference_landmark_decoding_type_type_e, respectively.

Change-Id: Ia3a4098213a712ffd1927838d7b9931d1edfcbd0
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>

10 files changed, 114 insertions, 62 deletions

diff --git a/meta-template/fld_mediapipe_192x192.json b/meta-template/fld_mediapipe_192x192.json
index 78e3b158..ee7f63c8 100644
--- a/meta-template/fld_mediapipe_192x192.json
+++ b/meta-template/fld_mediapipe_192x192.json
@@ -36,9 +36,9 @@
             {
                 "name" : "conv2d_20",
                 "index" : [-1, -1, -1, 1],
-                "landmark_type" : 0,
-                "landmark_coordinate" : 1,
-                "decoding_type" : 0,
+                "landmark_type" : "2D_SINGLE",
+                "landmark_coordinate" : "PIXEL",
+                "decoding_type" : "BYPASS",
                 "landmark_offset" : 3
             }
         ]
diff --git a/meta-template/fld_tweakcnn_128x128.json b/meta-template/fld_tweakcnn_128x128.json
index eaeeecda..0f2148c8 100644
--- a/meta-template/fld_tweakcnn_128x128.json
+++ b/meta-template/fld_tweakcnn_128x128.json
@@ -36,10 +36,10 @@
             {
                 "name" : "fanet8ss_inference/fully_connected_1/Sigmoid",
                 "index" : [-1, 1],
-                "landmark_type" : 0,
-                "landmark_coordinate" : 0,
+                "landmark_type" : "2D_SINGLE",
+                "landmark_coordinate" : "RATIO",
                 "landmark_offset" : 2,
-                "decoding_type" : 0
+                "decoding_type" : "BYPASS"
             }
         ]
     }
diff --git a/meta-template/pld_cpm_192x192.json b/meta-template/pld_cpm_192x192.json
index aa4ed690..a0bb6e6e 100644
--- a/meta-template/pld_cpm_192x192.json
+++ b/meta-template/pld_cpm_192x192.json
@@ -36,9 +36,9 @@
             {
                 "name" : "Convolutional_Pose_Machine/stage_5_out",
                 "index" : [-1, 1, 1, 1],
-                "landmark_type" : 0,
-                "landmark_coordinate" : 1,
-                "decoding_type" : 1,
+                "landmark_type" : "2D_SINGLE",
+                "landmark_coordinate" : "PIXEL",
+                "decoding_type" : "HEATMAP",
                 "decoding_info" :
                 {
                     "heatmap" :
diff --git a/meta-template/pld_mobilenet_v1_posenet_multi_257x257.json b/meta-template/pld_mobilenet_v1_posenet_multi_257x257.json
index 671e57cd..2ef057ab 100644
--- a/meta-template/pld_mobilenet_v1_posenet_multi_257x257.json
+++ b/meta-template/pld_mobilenet_v1_posenet_multi_257x257.json
@@ -36,9 +36,9 @@
             {
                 "name" : "MobilenetV1/heatmap_2/BiasAdd",
                 "index" : [-1, 1, 1, 1],
-                "landmark_type" : 1,
-                "landmark_coordinate" : 1,
-                "decoding_type" : 2,
+                "landmark_type" : "2D_MULTI",
+                "landmark_coordinate" : "PIXEL",
+                "decoding_type" : "HEATMAP_REFINE",
                 "decoding_info" :
                 {
                     "heatmap" :
diff --git a/mv_machine_learning/mv_inference/inference/include/OutputMetadata.h b/mv_machine_learning/mv_inference/inference/include/OutputMetadata.h
index 07488ec6..fe917ad4 100644
--- a/mv_machine_learning/mv_inference/inference/include/OutputMetadata.h
+++ b/mv_machine_learning/mv_inference/inference/include/OutputMetadata.h
@@ -287,29 +287,37 @@ namespace inference
 				int cIdx;
 				inference_tensor_shape_type_e shapeType;
 				float nmsRadius;
+				HeatMapInfo() = default;
+				~HeatMapInfo() = default;
 			};
 			HeatMapInfo heatMap;
+			DecodeInfo() = default;
+			~DecodeInfo() = default;
 		};
 	private:
 		std::string name;
 		DimInfo dimInfo;
-		int type; /**< 0: 2d-single, 1: 2d-multi, 2: 3-single */
+		inference_landmark_type_e type; /**< 0: 2D_SINGLE, 1: 2D_MULTI, 2: 3D_SINGLE */
 		int offset;
-		int coordinate; /**< 0: ratio, 1: pixel */
-		int decodingType; /**< 0: decoding  unnecessary,
-							1: decoding heatmap,
-							2: decoding heatmap with additional refine data */
+		inference_landmark_coorindate_type_e coordinate; /**< 0: RATIO, 1: PIXEL */
+		inference_landmark_decoding_type_e decodingType; /**< 0: decoding  unnecessary,
+															1: decoding heatmap,
+															2: decoding heatmap with refinement */
 		DecodeInfo decodingInfo;
 
+		std::map<std::string, inference_landmark_type_e> supportedLandmarkTypes;
+		std::map<std::string, inference_landmark_coorindate_type_e> supportedLandmarkCoordinateTypes;
+		std::map<std::string, inference_landmark_decoding_type_e> supportedLandmarkDecodingTypes;
+
 	public:
-		Landmark() = default;
+		Landmark();
 		~Landmark() = default;
 		std::string GetName() { return name; }
 		DimInfo GetDimInfo() { return dimInfo; }
-		int GetType();
+		inference_landmark_type_e GetType();
 		int GetOffset();
-		int GetCoordinate();
-		int GetDecodingType();
+		inference_landmark_coorindate_type_e GetCoordinate();
+		inference_landmark_decoding_type_e GetDecodingType();
 		DecodeInfo& GetDecodingInfo();
 
 		int ParseLandmark(JsonObject *root);
diff --git a/mv_machine_learning/mv_inference/inference/include/OutputMetadataTypes.h b/mv_machine_learning/mv_inference/inference/include/OutputMetadataTypes.h
index 523d0cb6..0a0aadce 100644
--- a/mv_machine_learning/mv_inference/inference/include/OutputMetadataTypes.h
+++ b/mv_machine_learning/mv_inference/inference/include/OutputMetadataTypes.h
@@ -52,6 +52,24 @@ namespace inference
 		INFERENCE_BOX_NMS_TYPE_NONE = -1,
 		INFERENCE_BOX_NMS_TYPE_STANDARD
 	} inference_box_nms_type_e;
+
+	// landmark
+	typedef enum {
+		INFERENCE_LANDMARK_TYPE_2D_SINGLE,
+		INFERENCE_LANDMARK_TYPE_2D_MULTI,
+		INFERENCE_LANDMARK_TYPE_3D_SINGLE
+	} inference_landmark_type_e;
+
+	typedef enum {
+		INFERENCE_LANDMARK_COORDINATE_TYPE_RATIO,
+		INFERENCE_LANDMARK_COORDINATE_TYPE_PIXEL
+	} inference_landmark_coorindate_type_e;
+
+	typedef enum {
+		INFERENCE_LANDMARK_DECODING_TYPE_BYPASS,
+		INFERENCE_LANDMARK_DECODING_TYPE_HEATMAP,
+		INFERENCE_LANDMARK_DECODING_TYPE_HEATMAP_REFINE
+	} inference_landmark_decoding_type_e;
 }
 }
 
diff --git a/mv_machine_learning/mv_inference/inference/src/Inference.cpp b/mv_machine_learning/mv_inference/inference/src/Inference.cpp
index 3db0155d..835bc6fa 100755
--- a/mv_machine_learning/mv_inference/inference/src/Inference.cpp
+++ b/mv_machine_learning/mv_inference/inference/src/Inference.cpp
@@ -1638,7 +1638,7 @@ namespace inference
 			int heatMapWidth = 0;
 			int heatMapHeight = 0;
 			int heatMapChannel = 0;
-			if (landmarkInfo.GetDecodingType() != 0) {
+			if (landmarkInfo.GetDecodingType() != INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 				heatMapWidth = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[landmarkInfo.GetDecodingInfo().heatMap.wIdx];
 				heatMapHeight = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[landmarkInfo.GetDecodingInfo().heatMap.hIdx];
 				heatMapChannel = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[landmarkInfo.GetDecodingInfo().heatMap.cIdx];
@@ -1646,7 +1646,7 @@ namespace inference
 
 			int number_of_landmarks = 0;
 			std::vector<int> channelIndexes = landmarkInfo.GetDimInfo().GetValidIndexAll();
-			if (landmarkInfo.GetDecodingType() == 0) {
+			if (landmarkInfo.GetDecodingType() == INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 				LOGI("landmark dim size: %zd and idx[0] is %d", channelIndexes.size(), channelIndexes[0]);
 				number_of_landmarks = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[channelIndexes[0]]
 									/ landmarkInfo.GetOffset();
@@ -1668,11 +1668,12 @@ namespace inference
 
 			float inputW = 1.f;
 			float inputH = 1.f;
-			if (landmarkInfo.GetCoordinate() == 1) {
+			if (landmarkInfo.GetCoordinate() == INFERENCE_LANDMARK_COORDINATE_TYPE_PIXEL) {
 				inputW = static_cast<float>(mMetadata.GetInputMeta().layer.begin()->second.GetWidth());
 				inputH = static_cast<float>(mMetadata.GetInputMeta().layer.begin()->second.GetHeight());
 			}
-			float thresRadius = landmarkInfo.GetType() == 0 ? 0.0 : outputMeta.GetLandmark().GetDecodingInfo().heatMap.nmsRadius;
+			float thresRadius = landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ? 0.0 :
+														outputMeta.GetLandmark().GetDecodingInfo().heatMap.nmsRadius;
 			poseDecoder.decode(inputW, inputH, thresRadius);
 
 			for (int landmarkIndex = 0; landmarkIndex < number_of_landmarks; landmarkIndex++) {
@@ -1730,7 +1731,7 @@ namespace inference
 			int heatMapWidth = 0;
 			int heatMapHeight = 0;
 			int heatMapChannel = 0;
-			if (landmarkInfo.GetDecodingType() != 0) {
+			if (landmarkInfo.GetDecodingType() != INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 				heatMapWidth = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[landmarkInfo.GetDecodingInfo().heatMap.wIdx];
 				heatMapHeight = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[landmarkInfo.GetDecodingInfo().heatMap.hIdx];
 				heatMapChannel = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[landmarkInfo.GetDecodingInfo().heatMap.cIdx];
@@ -1745,14 +1746,15 @@ namespace inference
 					return MEDIA_VISION_ERROR_INTERNAL;
 				}
 				// 2d+single or 2d+multi or 3d+single or 3d+multi
-				int defaultNumberOfPose = (landmarkInfo.GetType() == 0 || landmarkInfo.GetType() == 2) ? 1 : MAX_NUMBER_OF_POSE;
+				int defaultNumberOfPose = (landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ||
+										   landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_3D_SINGLE) ? 1 : MAX_NUMBER_OF_POSE;
 				std::vector<int> channelIndexes = landmarkInfo.GetDimInfo().GetValidIndexAll();
 
-				// In case of DecodingType == 0,
+				// If INFERENCE_LANDMARK_DECODING_TYPE_BYPASS,
 				// the landmarkChannel is guessed from the shape of the landmark output tensor.
-				// Otherwise, decoding heatmap, it is guessed from the heatMapChannel.
+				// Otherwise, it is guessed from the heatMapChannel.
 				int landmarkChannel = 0;
-				if (landmarkInfo.GetDecodingType() == 0) {
+				if (landmarkInfo.GetDecodingType() == INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 					landmarkChannel = mOutputLayerProperty.layers[landmarkInfo.GetName()].shape[channelIndexes[0]]
 										/ landmarkInfo.GetOffset();
 				} else {
@@ -1783,8 +1785,9 @@ namespace inference
 
 			float inputW = 1.f;
 			float inputH = 1.f;
-			float thresRadius = landmarkInfo.GetType() == 0 ? 0.0 : outputMeta.GetLandmark().GetDecodingInfo().heatMap.nmsRadius;
-			if (landmarkInfo.GetCoordinate() == 1) {
+			float thresRadius = landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ? 0.0 :
+														outputMeta.GetLandmark().GetDecodingInfo().heatMap.nmsRadius;
+			if (landmarkInfo.GetCoordinate() == INFERENCE_LANDMARK_COORDINATE_TYPE_PIXEL) {
 				inputW = static_cast<float>(mMetadata.GetInputMeta().layer.begin()->second.GetWidth());
 				inputH = static_cast<float>(mMetadata.GetInputMeta().layer.begin()->second.GetHeight());
 			}
diff --git a/mv_machine_learning/mv_inference/inference/src/OutputMetadata.cpp b/mv_machine_learning/mv_inference/inference/src/OutputMetadata.cpp
index 84c6c6d4..968bea38 100755
--- a/mv_machine_learning/mv_inference/inference/src/OutputMetadata.cpp
+++ b/mv_machine_learning/mv_inference/inference/src/OutputMetadata.cpp
@@ -706,6 +706,28 @@ namespace inference
 		return parsed;
 	}
 
+	Landmark::Landmark() :
+			name(),
+			dimInfo(),
+			type(INFERENCE_LANDMARK_TYPE_2D_SINGLE),
+			offset(),
+			coordinate(INFERENCE_LANDMARK_COORDINATE_TYPE_RATIO),
+			decodingType(INFERENCE_LANDMARK_DECODING_TYPE_BYPASS),
+			decodingInfo()
+
+	{
+		supportedLandmarkTypes.insert({"2D_SINGLE", INFERENCE_LANDMARK_TYPE_2D_SINGLE});
+		supportedLandmarkTypes.insert({"2D_MULTI",  INFERENCE_LANDMARK_TYPE_2D_MULTI});
+		supportedLandmarkTypes.insert({"3D_SINGLE", INFERENCE_LANDMARK_TYPE_3D_SINGLE});
+
+		supportedLandmarkCoordinateTypes.insert({"RATIO", INFERENCE_LANDMARK_COORDINATE_TYPE_RATIO});
+		supportedLandmarkCoordinateTypes.insert({"PIXEL", INFERENCE_LANDMARK_COORDINATE_TYPE_PIXEL});
+
+		supportedLandmarkDecodingTypes.insert({"BYPASS", INFERENCE_LANDMARK_DECODING_TYPE_BYPASS});
+		supportedLandmarkDecodingTypes.insert({"HEATMAP", INFERENCE_LANDMARK_DECODING_TYPE_HEATMAP});
+		supportedLandmarkDecodingTypes.insert({"HEATMAP_REFINE", INFERENCE_LANDMARK_DECODING_TYPE_HEATMAP_REFINE});
+	}
+
 	int Landmark::ParseLandmark(JsonObject *root)
 	{
 		// box
@@ -730,24 +752,24 @@ namespace inference
 					dimInfo.SetValidIndex(elem2);
 			}
 
-			type = static_cast<int>(json_object_get_int_member(pObject, "landmark_type"));
-			LOGI("landmark type: %d", type);
+			try {
+				type = OutputMetadata::GetSupportedType(pObject, "landmark_type", supportedLandmarkTypes);
+				coordinate = OutputMetadata::GetSupportedType(pObject, "landmark_coordinate", supportedLandmarkCoordinateTypes);
+				decodingType = OutputMetadata::GetSupportedType(pObject, "decoding_type", supportedLandmarkDecodingTypes);
+			} catch (const std::exception& e) {
+				LOGE("Invalid %s", e.what());
+				return MEDIA_VISION_ERROR_INVALID_OPERATION;
+			}
 
 			offset = static_cast<int>(json_object_get_int_member(pObject, "landmark_offset"));
 			LOGI("landmark offset: %d", offset);
-
-			coordinate = static_cast<int>(json_object_get_int_member(pObject, "landmark_coordinate"));
-			LOGI("landmark coordinate: %d", coordinate);
-
-			decodingType = static_cast<int>(json_object_get_int_member(pObject, "decoding_type"));
-			LOGI("landmark decodeing type: %d", decodingType);
 		}
 
 		LOGI("LEAVE");
 		return MEDIA_VISION_ERROR_NONE;
 	}
 
-	int Landmark::GetType()
+	inference_landmark_type_e Landmark::GetType()
 	{
 		return type;
 	}
@@ -757,12 +779,12 @@ namespace inference
 		return offset;
 	}
 
-	int Landmark::GetCoordinate()
+	inference_landmark_coorindate_type_e Landmark::GetCoordinate()
 	{
 		return coordinate;
 	}
 
-	int Landmark::GetDecodingType()
+	inference_landmark_decoding_type_e Landmark::GetDecodingType()
 	{
 		return decodingType;
 	}
@@ -1046,8 +1068,7 @@ namespace inference
 		}
 
 		if (!landmark.GetName().empty()) {
-			if (landmark.GetDecodingType() == 1 ||
-				landmark.GetDecodingType() == 2) {
+			if (landmark.GetDecodingType() != INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 				ret = ParseLandmarkDecodeInfo(root);
 				if (ret != MEDIA_VISION_ERROR_NONE) {
 					LOGE("Fail to GetLandmarkDecodeInfo[%d]", ret);
@@ -1055,7 +1076,7 @@ namespace inference
 				}
 			}
 
-			if (landmark.GetDecodingType() == 2) {// landmark.decodingType == 2
+			if (landmark.GetDecodingType() == INFERENCE_LANDMARK_DECODING_TYPE_HEATMAP_REFINE) {
 				ret = ParseOffset(root);
 				if (ret != MEDIA_VISION_ERROR_NONE) {
 					LOGE("Fail to GetOffsetVector[%d]", ret);
diff --git a/mv_machine_learning/mv_inference/inference/src/PoseDecoder.cpp b/mv_machine_learning/mv_inference/inference/src/PoseDecoder.cpp
index 77116735..9798dfcf 100644
--- a/mv_machine_learning/mv_inference/inference/src/PoseDecoder.cpp
+++ b/mv_machine_learning/mv_inference/inference/src/PoseDecoder.cpp
@@ -52,12 +52,13 @@ namespace inference
 
 		Landmark& landmarkInfo = mMeta.GetLandmark();
 
-		if (landmarkInfo.GetType() < 0 || landmarkInfo.GetType() >= 3) {
+		if (landmarkInfo.GetType() < INFERENCE_LANDMARK_TYPE_2D_SINGLE ||
+			landmarkInfo.GetType() > INFERENCE_LANDMARK_TYPE_3D_SINGLE) {
 			LOGE("Not supported landmark type");
 			return MEDIA_VISION_ERROR_INVALID_OPERATION;
 		}
 
-		if (landmarkInfo.GetDecodingType() == 0) {
+		if (landmarkInfo.GetDecodingType() == INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 			LOGI("Skip init");
 			return MEDIA_VISION_ERROR_NONE;
 		}
@@ -71,8 +72,8 @@ namespace inference
 
 		mCandidates.clear();
 
-		if (landmarkInfo.GetType() == 0 ||
-			landmarkInfo.GetType() == 2) {
+		if (landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ||
+			landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_3D_SINGLE) {
 			mCandidates.resize(mHeatMapChannel);
 		}
 
@@ -90,8 +91,8 @@ namespace inference
 					if (score < scoreInfo.GetThresHold())
 						continue;
 
-					if (landmarkInfo.GetType() == 0 ||
-						landmarkInfo.GetType() == 2) {
+					if (landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ||
+						landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_3D_SINGLE) {
 						if (score <= candidate->score)
 							continue;
 
@@ -245,25 +246,25 @@ namespace inference
 		Landmark& landmarkInfo = mMeta.GetLandmark();
 		ScoreInfo& scoreInfo = mMeta.GetScore();
 
-		if (landmarkInfo.GetType() == 0 ||
-			landmarkInfo.GetType() == 2) { // single pose
+		if (landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ||
+			landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_3D_SINGLE) {
 			mPoseLandmarks.resize(1);
 
-			if (landmarkInfo.GetDecodingType() == 0) { // direct decoding
+			if (landmarkInfo.GetDecodingType() == INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 				mPoseLandmarks[0].landmarks.resize(mNumberOfLandmarks);
-			} else { // heatmap decoding
+			} else {
 				mPoseLandmarks[0].landmarks.resize(mHeatMapChannel);
 			}
 		}
 
-		if (landmarkInfo.GetDecodingType() != 0) { // heatmap decoding
+		if (landmarkInfo.GetDecodingType() != INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 			while (!mCandidates.empty()) {
 
 				LandmarkPoint &root = mCandidates.front();
 
 				getIndexToPos(root, scaleWidth, scaleHeight);
 
-				if (landmarkInfo.GetType() == 0) {
+				if (landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE) {
 					root.valid = true;
 					mPoseLandmarks[0].landmarks[root.id] = root;
 					mPoseLandmarks[0].score += root.score;
@@ -330,8 +331,9 @@ namespace inference
 				}
 			}
 
-			int landmarkOffset = (landmarkInfo.GetType() == 0 || landmarkInfo.GetType() == 1) ? 2 : 3;
-			if (landmarkInfo.GetDecodingType() == 0) {
+			int landmarkOffset = (landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_SINGLE ||
+								  landmarkInfo.GetType() == INFERENCE_LANDMARK_TYPE_2D_MULTI) ? 2 : 3;
+			if (landmarkInfo.GetDecodingType() == INFERENCE_LANDMARK_DECODING_TYPE_BYPASS) {
 				landmarkOffset = landmarkInfo.GetOffset();
 			}
 			for (int idx = 0; idx < mNumberOfLandmarks; ++idx) {
diff --git a/packaging/capi-media-vision.spec b/packaging/capi-media-vision.spec
index 7aeb8a11..a1ca7071 100644
--- a/packaging/capi-media-vision.spec
+++ b/packaging/capi-media-vision.spec
@@ -1,7 +1,7 @@
 Name:        capi-media-vision
 Summary:     Media Vision library for Tizen Native API
-Version:     0.8.12
-Release:     2
+Version:     0.8.13
+Release:     0
 Group:       Multimedia/Framework
 License:     Apache-2.0 and BSD-3-Clause
 Source0:     %{name}-%{version}.tar.gz