Migrate ObjectDetection to common CV utils

Author: benITo47

packages/react-native-executorch/common/rnexecutorch/host_objects/JsiConversions.h

@@ -446,13 +446,8 @@ inline jsi::Value getJsiValue(
   jsi::Array array(runtime, detections.size());
   for (std::size_t i = 0; i < detections.size(); ++i) {
     jsi::Object detection(runtime);
-    jsi::Object bbox(runtime);
-    bbox.setProperty(runtime, "x1", detections[i].x1);
-    bbox.setProperty(runtime, "y1", detections[i].y1);
-    bbox.setProperty(runtime, "x2", detections[i].x2);
-    bbox.setProperty(runtime, "y2", detections[i].y2);
-
-    detection.setProperty(runtime, "bbox", bbox);
+    detection.setProperty(runtime, "bbox",
+                          getJsiValue(detections[i].bbox, runtime));
     detection.setProperty(
         runtime, "label",
         jsi::String::createFromUtf8(runtime, detections[i].label));
@@ -462,10 +457,10 @@ inline jsi::Value getJsiValue(
   return array;
 }
 
-inline jsi::Value getJsiValue(
-    const std::vector<models::instance_segmentation::types::InstanceMask>
-        &instances,
-    jsi::Runtime &runtime) {
+inline jsi::Value
+getJsiValue(const std::vector<models::instance_segmentation::types::Instance>
+                &instances,
+            jsi::Runtime &runtime) {
   jsi::Array array(runtime, instances.size());
   for (std::size_t i = 0; i < instances.size(); ++i) {
     jsi::Object instance(runtime);

packages/react-native-executorch/common/rnexecutorch/models/instance_segmentation/BaseInstanceSegmentation.cpp

@@ -90,7 +90,7 @@ cv::Mat BaseInstanceSegmentation::processMaskFromLogits(
   return finalBinaryMat;
 }
 
-std::optional<types::InstanceMask> BaseInstanceSegmentation::processDetection(
+std::optional<types::Instance> BaseInstanceSegmentation::processDetection(
     int32_t detectionIndex, const float *bboxData, const float *scoresData,
     const float *maskData, int32_t maskH, int32_t maskW,
     cv::Size modelInputSize, cv::Size originalSize, float widthRatio,
@@ -140,13 +140,13 @@ std::optional<types::InstanceMask> BaseInstanceSegmentation::processDetection(
   std::vector<uint8_t> finalMask(finalBinaryMat.data,
                                  finalBinaryMat.data + finalBinaryMat.total());
 
-  return types::InstanceMask(
+  return types::Instance(
       utils::computer_vision::BBox{origX1, origY1, origX2, origY2},
       std::move(finalMask), finalMaskWidth, finalMaskHeight,
       static_cast<int32_t>(labelIdx), score, i);
 }
 
-std::vector<types::InstanceMask> BaseInstanceSegmentation::postprocess(
+std::vector<types::Instance> BaseInstanceSegmentation::postprocess(
     const std::vector<EValue> &tensors, cv::Size originalSize,
     cv::Size modelInputSize, double confidenceThreshold, double iouThreshold,
     int32_t maxInstances, const std::vector<int32_t> &classIndices,
@@ -175,7 +175,7 @@ std::vector<types::InstanceMask> BaseInstanceSegmentation::postprocess(
     allowedClasses.insert(classIndices.begin(), classIndices.end());
   }
 
-  std::vector<types::InstanceMask> instances;
+  std::vector<types::Instance> instances;
 
   size_t numTensors = tensors.size();
   if (numTensors != 3) {
@@ -231,7 +231,7 @@ std::vector<types::InstanceMask> BaseInstanceSegmentation::postprocess(
   return instances;
 }
 
-std::vector<types::InstanceMask> BaseInstanceSegmentation::generate(
+std::vector<types::Instance> BaseInstanceSegmentation::generate(
     std::string imageSource, double confidenceThreshold, double iouThreshold,
     int32_t maxInstances, std::vector<int32_t> classIndices,
     bool returnMaskAtOriginalResolution, std::string methodName) {

packages/react-native-executorch/common/rnexecutorch/models/instance_segmentation/BaseInstanceSegmentation.h

@@ -22,14 +22,14 @@ class BaseInstanceSegmentation : public BaseModel {
                            std::vector<float> normStd, bool applyNMS,
                            std::shared_ptr<react::CallInvoker> callInvoker);
 
-  [[nodiscard("Registered non-void function")]] std::vector<types::InstanceMask>
+  [[nodiscard("Registered non-void function")]] std::vector<types::Instance>
   generate(std::string imageSource, double confidenceThreshold,
            double iouThreshold, int32_t maxInstances,
            std::vector<int32_t> classIndices,
            bool returnMaskAtOriginalResolution, std::string methodName);
 
 private:
-  std::vector<types::InstanceMask>
+  std::vector<types::Instance>
   postprocess(const std::vector<EValue> &tensors, cv::Size originalSize,
               cv::Size modelInputSize, double confidenceThreshold,
               double iouThreshold, int32_t maxInstances,
@@ -43,7 +43,7 @@ class BaseInstanceSegmentation : public BaseModel {
                                 float origX1, float origY1, bool warpToOriginal,
                                 int32_t &outWidth, int32_t &outHeight);
 
-  std::optional<types::InstanceMask>
+  std::optional<types::Instance>
   processDetection(int32_t detectionIndex, const float *bboxData,
                    const float *scoresData, const float *maskData,
                    int32_t maskH, int32_t maskW, cv::Size modelInputSize,

packages/react-native-executorch/common/rnexecutorch/models/instance_segmentation/Types.h

@@ -12,7 +12,7 @@ namespace rnexecutorch::models::instance_segmentation::types {
  * Contains bounding box coordinates, binary segmentation mask, class label,
  * confidence score, and a unique instance identifier.
  */
-struct InstanceMask {
+struct Instance {
   utils::computer_vision::BBox bbox; ///< Bounding box coordinates
   std::vector<uint8_t> mask;         ///< Binary mask (0 or 1) for the instance
   int32_t maskWidth;                 ///< Width of the mask array
@@ -21,15 +21,10 @@ struct InstanceMask {
   float score;                       ///< Confidence score [0, 1]
   int32_t instanceId;                ///< Unique identifier for this instance
 
-  InstanceMask() = default;
-  InstanceMask(const InstanceMask &) = default;
-  InstanceMask(InstanceMask &&) = default;
-  InstanceMask &operator=(const InstanceMask &) = default;
-  InstanceMask &operator=(InstanceMask &&) = default;
-
-  InstanceMask(utils::computer_vision::BBox bbox, std::vector<uint8_t> mask,
-               int32_t maskWidth, int32_t maskHeight, int32_t classIndex,
-               float score, int32_t instanceId)
+  Instance() = default;
+  Instance(utils::computer_vision::BBox bbox, std::vector<uint8_t> mask,
+           int32_t maskWidth, int32_t maskHeight, int32_t classIndex,
+           float score, int32_t instanceId)
       : bbox(bbox), mask(std::move(mask)), maskWidth(maskWidth),
         maskHeight(maskHeight), classIndex(classIndex), score(score),
         instanceId(instanceId) {}

packages/react-native-executorch/common/rnexecutorch/models/object_detection/ObjectDetection.cpp

@@ -1,11 +1,13 @@
 #include "ObjectDetection.h"
+#include "Constants.h"
 
 #include <rnexecutorch/Error.h>
 #include <rnexecutorch/ErrorCodes.h>
 #include <rnexecutorch/Log.h>
 #include <rnexecutorch/data_processing/ImageProcessing.h>
 #include <rnexecutorch/host_objects/JsiConversions.h>
 #include <rnexecutorch/utils/FrameProcessor.h>
+#include <rnexecutorch/utils/computer_vision/Processing.h>
 
 namespace rnexecutorch::models::object_detection {
 
@@ -119,10 +121,13 @@ ObjectDetection::postprocess(const std::vector<EValue> &tensors,
               " exceeds labelNames size " + std::to_string(labelNames_.size()) +
               ". Ensure the labelMap covers all model output classes.");
     }
-    detections.emplace_back(x1, y1, x2, y2, labelNames_[labelIdx], scores[i]);
+    detections.emplace_back(utils::computer_vision::BBox{x1, y1, x2, y2},
+                            labelNames_[labelIdx],
+                            static_cast<int32_t>(labelIdx), scores[i]);
   }
 
-  return utils::nonMaxSuppression(detections);
+  return utils::computer_vision::nonMaxSuppression(detections,
+                                                   constants::IOU_THRESHOLD);
 }
 
 std::vector<types::Detection>

packages/react-native-executorch/common/rnexecutorch/models/object_detection/ObjectDetection.h

@@ -8,7 +8,6 @@
 #include "Types.h"
 #include "rnexecutorch/metaprogramming/ConstructorHelpers.h"
 #include <rnexecutorch/models/VisionModel.h>
-#include <rnexecutorch/models/object_detection/Utils.h>
 
 namespace rnexecutorch {
 namespace models::object_detection {

packages/react-native-executorch/common/rnexecutorch/models/object_detection/Types.h

@@ -1,15 +1,21 @@
 #pragma once
 
+#include <cstdint>
+#include <rnexecutorch/utils/computer_vision/Types.h>
 #include <string>
 
 namespace rnexecutorch::models::object_detection::types {
 struct Detection {
-  float x1;
-  float y1;
-  float x2;
-  float y2;
+  utils::computer_vision::BBox bbox;
   std::string label;
+  int32_t classIndex;
   float score;
+
+  Detection() = default;
+  Detection(utils::computer_vision::BBox bbox, std::string label,
+            int32_t classIndex, float score)
+      : bbox(bbox), label(std::move(label)), classIndex(classIndex),
+        score(score) {}
 };
 
-} // namespace rnexecutorch::models::object_detection::types
+} // namespace rnexecutorch::models::object_detection::types