Merge branch 'feature/cpp_refactoring' into rhecker/model_api_refactor_classification

Author: RHeckerIntel

examples/cpp/main.cpp

@@ -4,7 +4,6 @@
  */
 
 #include <stddef.h>
-#include <tasks/classification.h>
 #include <tasks/detection.h>
 #include <tasks/results.h>
 
@@ -18,25 +17,34 @@
 #include <stdexcept>
 #include <string>
 
-int main(int argc, char* argv[]) {
+int main(int argc, char* argv[]) try {
     if (argc != 3) {
         throw std::runtime_error(std::string{"Usage: "} + argv[0] + " <path_to_model> <path_to_image>");
     }
 
     cv::Mat image = cv::imread(argv[2]);
-    //cv::cvtColor(image, image, cv::COLOR_BGR2RGB);
+    cv::cvtColor(image, image, cv::COLOR_BGR2RGB);
 
     if (!image.data) {
         throw std::runtime_error{"Failed to read the image"};
     }
 
-    //// Instantiate Object Detection model
-    auto model = Classification::load(argv[1]);  // works with SSD models. Download it using Python Model API
+    // Instantiate Object Detection model
+    auto model = DetectionModel::load(argv[1], {});  // works with SSD models. Download it using Python Model API
 
-    //// Run the inference
+    // Run the inference
     auto result = model.infer(image);
 
-    //// Process detections
-    std::cout << result << std::endl;
-    std::cout << "expected: " << std::endl << "1 (bicycle): 0.825, 11 (dog): 0.873, 14 (person): 0.824, [0], [0], [0]" << std::endl;
+    // Process detections
+    for (auto& obj : result.objects) {
+        std::cout << " " << std::left << std::setw(9) << obj.label << " | " << std::setw(10) << obj.confidence << " | "
+                  << std::setw(4) << int(obj.x) << " | " << std::setw(4) << int(obj.y) << " | " << std::setw(4)
+                  << int(obj.x + obj.width) << " | " << std::setw(4) << int(obj.y + obj.height) << "\n";
+    }
+} catch (const std::exception& error) {
+    std::cerr << error.what() << '\n';
+    return 1;
+} catch (...) {
+    std::cerr << "Non-exception object thrown\n";
+    return 1;
 }

src/cpp/include/tasks/detection/ssd.h

@@ -9,6 +9,7 @@
 
 #include "adapters/inference_adapter.h"
 #include "tasks/results.h"
+#include "utils/config.h"
 #include "utils/preprocessing.h"
 
 enum SSDOutputMode { single, multi };
@@ -29,20 +30,8 @@ class SSD {
 
     SSD(std::shared_ptr<InferenceAdapter> adapter, cv::Size input_shape) : adapter(adapter), input_shape(input_shape) {
         auto config = adapter->getModelConfig();
-        {
-            auto iter = config.find("labels");
-            if (iter != config.end()) {
-                labels = iter->second.as<std::vector<std::string>>();
-            } else {
-                std::cout << "could not find labels from model config" << std::endl;
-            }
-        }
-        {
-            auto iter = config.find("confidence_threshold");
-            if (iter != config.end()) {
-                confidence_threshold = iter->second.as<float>();
-            }
-        }
+        labels = utils::get_from_any_maps("labels", config, {}, labels);
+        confidence_threshold = utils::get_from_any_maps("confidence_threshold", config, {}, confidence_threshold);
     }
     std::map<std::string, ov::Tensor> preprocess(cv::Mat);
     DetectionResult postprocess(InferenceResult& infResult);

src/cpp/include/tasks/instance_segmentation.h

@@ -10,6 +10,7 @@
 
 #include "adapters/inference_adapter.h"
 #include "tasks/results.h"
+#include "utils/config.h"
 #include "utils/vision_pipeline.h"
 
 class InstanceSegmentation {
@@ -30,20 +31,10 @@ class InstanceSegmentation {
             });
 
         auto config = adapter->getModelConfig();
-        auto iter = config.find("labels");
-        if (iter != config.end()) {
-            labels = iter->second.as<std::vector<std::string>>();
-        } else {
-            std::cout << "could not find labels from model config" << std::endl;
-        }
-
-        {
-            auto iter = config.find("confidence_threshold");
-            if (iter != config.end()) {
-                confidence_threshold = iter->second.as<float>();
-            }
-        }
+        labels = utils::get_from_any_maps("labels", config, {}, labels);
+        confidence_threshold = utils::get_from_any_maps("confidence_threshold", config, {}, confidence_threshold);
     }
+
     static cv::Size serialize(std::shared_ptr<ov::Model>& ov_model);
     static InstanceSegmentation load(const std::string& model_path);
 

src/cpp/include/tasks/semantic_segmentation.h

@@ -9,6 +9,7 @@
 
 #include "adapters/inference_adapter.h"
 #include "tasks/results.h"
+#include "utils/config.h"
 #include "utils/preprocessing.h"
 #include "utils/vision_pipeline.h"
 
@@ -27,26 +28,9 @@ class SemanticSegmentation {
             });
 
         auto config = adapter->getModelConfig();
-        auto iter = config.find("labels");
-        if (iter != config.end()) {
-            labels = iter->second.as<std::vector<std::string>>();
-        } else {
-            std::cout << "could not find labels from model config" << std::endl;
-        }
-
-        {
-            auto iter = config.find("soft_threshold");
-            if (iter != config.end()) {
-                soft_threshold = iter->second.as<float>();
-            }
-        }
-
-        {
-            auto iter = config.find("blur_strength");
-            if (iter != config.end()) {
-                blur_strength = iter->second.as<int>();
-            }
-        }
+        labels = utils::get_from_any_maps("labels", config, {}, labels);
+        soft_threshold = utils::get_from_any_maps("soft_threshold", config, {}, soft_threshold);
+        blur_strength = utils::get_from_any_maps("blur_strength", config, {}, blur_strength);
     }
 
     static cv::Size serialize(std::shared_ptr<ov::Model>& ov_model);

src/cpp/src/tasks/detection/ssd.cpp

@@ -148,15 +148,6 @@ void SSD::prepareMultipleOutputs(std::shared_ptr<ov::Model> ov_model) {
     for (auto& name : output_names) {
         std::cout << "output name: " << name << std::endl;
     }
-
-    // ov::preprocess::PrePostProcessor ppp(ov_model);
-
-    // for (const auto& output_name : output_names) {
-    //     if (output_name != "labels") { //TODO: Discover why this isnt needed in original?
-    //         ppp.output(output_name).tensor().set_element_type(ov::element::f32);
-    //     }
-    // }
-    // ov_model = ppp.build();
 }
 
 std::vector<std::string> SSD::filterOutXai(const std::vector<std::string>& names) {

tests/cpp/public_scope.json

@@ -0,0 +1,100 @@
+[
+  {
+    "name": "otx_models/Lite-hrnet-18.xml",
+    "type": "SegmentationModel",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "0: 0.537, 1: 0.463, [426,640,2], [0], [0]; object: 0.675, 508, object: 0.527, 65, object: 0.507, 18, object: 0.624, 144, object: 0.538, 67, object: 0.507, 15, object: 0.518, 41, object: 0.507, 8, object: 0.505, 14, object: 0.885, 2138, "
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/Lite-hrnet-18_mod2.xml",
+    "type": "SegmentationModel",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "0: 0.992, 1: 0.008, [426,640,2], [0], [0]; object: 0.555, 112, object: 0.506, 17, object: 0.555, 154, object: 0.511, 19, object: 0.514, 52, "
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/segmentation_model_with_xai_head.xml",
+    "type": "SegmentationModel",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "0: 1.000, 1: 0.000, [426,640,3], [426,640,3], [1,600,1,1]; backpack: 0.505, 2, "
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/maskrcnn_model_with_xai_head.xml",
+    "type": "MaskRCNNModel",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "61, 277, 358, 382, 17 (horse): 0.998, 18312, RotatedRect: 212.000 327.000 290.000 100.000 0.000; 1, 14, 162, 321, 2 (car): 0.994, 25867, RotatedRect: 54.067 173.034 285.208 156.889 61.996; 327, 96, 341, 134, 1 (bicycle): 0.930, 279, RotatedRect: 333.500 114.000 36.000 13.000 90.000; 460, 106, 493, 148, 1 (bicycle): 0.898, 786, RotatedRect: 476.284 126.621 27.308 45.993 19.179; 294, 93, 315, 153, 1 (bicycle): 0.869, 789, RotatedRect: 304.000 124.000 58.000 18.000 90.000; 278, 109, 290, 152, 1 (bicycle): 0.817, 355, RotatedRect: 283.500 130.000 42.000 11.000 90.000; 4, 4, 102, 191, 2 (car): 0.701, 9658, RotatedRect: 51.806 97.259 184.445 95.281 89.246; 270, 93, 290, 152, 1 (bicycle): 0.660, 723, RotatedRect: 280.500 122.500 17.000 59.000 0.000; 322, 114, 343, 152, 18 (sheep): 0.520, 298, RotatedRect: 332.000 133.000 34.000 14.000 90.000; 4; [1,1280,1,1]; "
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/is_resnet50_maskrcnn_coco_reduced.xml",
+    "type": "MaskRCNNModel",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "59, 277, 360, 380, 16 (horse): 0.999, 19053, RotatedRect: 210.000 327.500 101.000 296.000 90.000; 2, 9, 162, 318, 2 (car): 0.999, 31153, RotatedRect: 82.086 163.312 307.394 156.997 89.669; 294, 94, 316, 153, 1 (bicycle): 0.985, 840, RotatedRect: 305.000 123.500 59.000 18.000 90.000; 326, 97, 341, 136, 1 (bicycle): 0.974, 397, RotatedRect: 332.500 116.000 38.000 13.000 90.000; 461, 105, 493, 150, 1 (bicycle): 0.918, 846, RotatedRect: 476.052 126.972 27.619 47.834 16.928; 350, 92, 386, 149, 1 (bicycle): 0.807, 1458, RotatedRect: 369.319 119.891 54.848 34.230 82.405; 279, 110, 291, 146, 1 (bicycle): 0.788, 312, RotatedRect: 284.000 127.500 35.000 10.000 90.000; 0; [0]; horse: 0.999, 668; car: 0.999, 782; bicycle: 0.985, 127; bicycle: 0.974, 87; bicycle: 0.918, 122; bicycle: 0.807, 140; bicycle: 0.788, 79; "
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/det_mobilenetv2_atss_bccd.xml",
+    "type": "DetectionModel",
+    "test_data": [
+      {
+        "image": "BloodImage_00007.jpg",
+        "reference": [
+          "494, 159, 637, 308, 2 (WBC): 0.697; 28, 139, 135, 228, 1 (RBC): 0.628; 535, 375, 638, 479, 1 (RBC): 0.524; 513, 8, 633, 152, 1 (RBC): 0.430; 21, 291, 143, 399, 1 (RBC): 0.422; 196, 86, 410, 286, 1 (RBC): 0.422; [0]; [0]"
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/detection_model_with_xai_head.xml",
+    "type": "DetectionModel",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "61, 277, 355, 379, 1 (person): 0.364; 461, 105, 495, 149, 1 (person): 0.305; [1,2,6,8]; [1,320,1,1]"
+        ]
+      }
+    ]
+  },
+  {
+    "name": "otx_models/detection_model_with_xai_head.xml",
+    "type": "DetectionModel",
+    "tiler": "DetectionTiler",
+    "input_res": "(3500,3500)",
+    "test_data": [
+      {
+        "image": "coco128/images/train2017/000000000074.jpg",
+        "reference": [
+          "336, 2275, 1944, 3114, 1 (person): 0.361; 2523, 862, 2709, 1224, 1 (person): 0.313; [1,2,35,46]; [1,320,1,1]"
+        ]
+      }
+    ]
+  }
+]

tests/cpp/test_accuracy.cpp

@@ -103,13 +103,14 @@ TEST_P(ModelParameterizedTest, AccuracyTest) {
     } else if (data.type == "MaskRCNNModel") {
         GTEST_SKIP();
     } else if (data.type == "ClassificationModel") {
-        auto model = Classification::load(model_path);
-        for (auto& test_data : data.test_data) {
-            std::string image_path = DATA_DIR + '/' + test_data.image;
-            cv::Mat image = cv::imread(image_path);
-            auto result = model.infer(image);
-            EXPECT_EQ(std::string{result}, test_data.reference[0]);
-        }
+        GTEST_SKIP();
+        //auto model = Classification::load(model_path);
+        //for (auto& test_data : data.test_data) {
+        //    std::string image_path = DATA_DIR + '/' + test_data.image;
+        //    cv::Mat image = cv::imread(image_path);
+        //    auto result = model.infer(image);
+        //    EXPECT_EQ(std::string{result}, test_data.reference[0]);
+        //}
     } else {
         FAIL() << "No implementation for model type " << data.type;
     }