更新到HyperLPR3版本

cpp/test/nn_module/test_classification.cpp

@@ -0,0 +1,54 @@
+//
+// Created by tunm on 2023/2/11.
+//
+
+
+#include "basic_types.h"
+#include "../test_settings.h"
+#include "opencv2/opencv.hpp"
+#include "nn_implementation_module/classification/all.h"
+#include "basic_types.h"
+#include "utils.h"
+
+using namespace hyper;
+
+TEST_CASE("test_Classification", "[nn_cls]") {
+    PRINT_SPLIT_LINE
+    LOGD("[UnitTest]->Classification Model");
+
+    std::string model_path = GET_DATA("models/r2_mobile/litemodel_cls_96xh.mnn");
+
+    std::vector<std::string> predict_images_list = {
+            GET_DATA("images/align/1.jpg"),
+            GET_DATA("images/align/3.jpg"),
+            GET_DATA("images/align/5.jpg"),
+    };
+    std::vector<PlateColor> predict_results_cls = {
+            PlateColor::BLUE, PlateColor::YELLOW, PlateColor::GREEN,
+    };
+    std::vector<float> predict_results_confidence = {
+            0.9999293f, 0.8975975f, 0.9997952f
+    };
+
+    CHECK(predict_results_confidence.size() == predict_results_cls.size());
+    CHECK(predict_results_confidence.size() == predict_images_list.size());
+
+    ClassificationEngine clsEngine;
+    auto ret = clsEngine.Initialize(model_path, cv::Size_<int>(96, 96));
+    CHECK(ret == InferenceHelper::kRetOk);
+
+    SECTION("test_ClassificationModelPredict") {
+        for (int i = 0; i < predict_images_list.size(); ++i) {
+            cv::Mat img = cv::imread(predict_images_list[i]);
+            CHECK(!img.empty());
+            CHECK(img.cols == 96);
+            CHECK(img.rows == 96);
+            ret = clsEngine.Inference(img);
+            CHECK(ret == InferenceHelper::kRetOk);
+            CHECK(PlateColor(clsEngine.getMOutputColor()) == predict_results_cls[i]);
+            CHECK(clsEngine.getMOutputMaxConfidence() == Approx(predict_results_confidence[i]).epsilon(0.001));
+        }
+    }
+
+
+}

cpp/test/nn_module/test_detection.cpp

@@ -0,0 +1,46 @@
+//
+// Created by tunm on 2023/2/11.
+//
+#include "basic_types.h"
+#include "../test_settings.h"
+#include "opencv2/opencv.hpp"
+#include "nn_implementation_module/detect/all.h"
+#include "basic_types.h"
+#include "utils.h"
+
+using namespace hyper;
+
+TEST_CASE("test_Detection", "[nn_detect]") {
+    PRINT_SPLIT_LINE
+    LOGD("[UnitTest]->Detect Model");
+
+    std::string b_model_path = GET_DATA("models/r2_mobile/b320_backbone_h.mnn");
+    std::string h_model_path = GET_DATA("models/r2_mobile/b320_header_h.mnn");
+
+    cv::Mat test_image_1 = cv::imread(GET_DATA("images/pre.jpg"));
+    CHECK(test_image_1.cols == 320);
+    CHECK(test_image_1.rows == 320);
+
+    SECTION("test_SplitDetectionSplitModel") {
+        LOGD("Detect Model SplitModel");
+        DetArch det;
+        auto ret = det.Initialize(b_model_path, h_model_path, 320, 1);
+        CHECK(ret == InferenceHelper::kRetOk);
+
+        det.Detection(test_image_1);
+        auto &result = det.m_results_;
+        CHECK(result.size() == 1);
+
+        auto &box = result[0];
+        cv::Rect2f proposal_box(cv::Point(153, 205), cv::Point(183, 215));
+        cv::Rect2f detected_box(cv::Point(box.x1, box.y1), cv::Point(box.x2, box.y2));
+        auto iou = boundBoxOverlap(proposal_box, detected_box);
+        CHECK(iou > 0.85);
+    }
+
+#if ENABLE_BENCHMARK_TEST
+    SECTION("test_DetectionBenchmark") {
+        LOGD("[UnitTest]->Detection Benchmark");
+    }
+#endif
+}

cpp/test/nn_module/test_recognition.cpp

@@ -0,0 +1,51 @@
+//
+// Created by tunm on 2023/2/11.
+//
+#include "basic_types.h"
+#include "../test_settings.h"
+#include "opencv2/opencv.hpp"
+#include "nn_implementation_module/recognition/all.h"
+#include "basic_types.h"
+#include "utils.h"
+
+using namespace hyper;
+
+TEST_CASE("test_Recognition", "[nn_rec]") {
+    PRINT_SPLIT_LINE
+    LOGD("[UnitTest]->Recognition Model");
+
+    std::string model_path = GET_DATA("models/r2_mobile/rpv3_mdict_160h.mnn");
+
+    std::vector<std::string> predict_images_list = {
+            GET_DATA("images/rec_crop/_0_津B6H920.jpg"),
+            GET_DATA("images/rec_crop/_1_皖KD01833.jpg"),
+            GET_DATA("images/rec_crop/_6_蒙B023H6.jpg"),
+            GET_DATA("images/rec_crop/_8_冀D5L690.jpg"),
+    };
+
+    std::vector<std::string> predict_results_code = {
+            "津B6H920", "皖KD01833", "蒙B023H6", "冀D5L690",
+    };
+
+    RecognitionEngine recEngine;
+    auto ret = recEngine.Initialize(model_path);
+    CHECK(ret == InferenceHelper::kRetOk);
+
+    SECTION("test_SplitDetectionSplitModel") {
+        LOGD("Rec Model RPV3");
+        for (int i = 0; i < predict_images_list.size(); ++i) {
+            cv::Mat img = cv::imread(predict_images_list[i]);
+            CHECK(!img.empty());
+            float wh_ratio = (float) img.cols / img.rows;
+            cv::Mat align_image_pad;
+            imagePadding(img, align_image_pad, wh_ratio, recEngine.getMInputImageSize());
+            TextLine line;
+            ret = recEngine.Inference(align_image_pad, line);
+            LOGD("%s -> %s", predict_results_code[i].c_str(), line.code.c_str());
+            CHECK(ret == InferenceHelper::kRetOk);
+            CHECK(strcmp(predict_results_code[i].c_str(), line.code.c_str()) == 0);
+        }
+
+    }
+
+}