potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-30 15:53:03 +04:00
Code Packages Releases Wiki Activity

Darknet cleanup

Browse Source
This commit is contained in:
Prasad Ayush Kumar
2026-04-30 14:43:11 +05:30
parent 0dbe5e7c35
commit 5af1571073
15 changed files with 255 additions and 2612 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/test/test_common.hpp
+18
View File
@@ -6,6 +6,7 @@
#define __OPENCV_TEST_COMMON_HPP__
#include "opencv2/dnn/utils/inference_engine.hpp"
#include <string>
#ifdef HAVE_OPENCL
#include "opencv2/core/ocl.hpp"
@@ -247,6 +248,23 @@ protected:
void runLayer(cv::Ptr<cv::dnn::Layer> layer, std::vector<cv::Mat> &inpBlobs, std::vector<cv::Mat> &outBlobs);
inline std::string getCurrentTestNameNoParams()
{
const ::testing::TestInfo* const test_info =
::testing::UnitTest::GetInstance()->current_test_info();
if (!test_info)
return std::string();
std::string suite = test_info->test_case_name();
std::string name = test_info->name();
const auto suite_slash = suite.find('/');
if (suite_slash != std::string::npos)
suite = suite.substr(0, suite_slash);
const auto name_slash = name.find('/');
if (name_slash != std::string::npos)
name = name.substr(0, name_slash);
return suite + "." + name;
}
} // namespace
#endif
modules/dnn/test/test_darknet_importer.cpp
+211 -867
View File
File diff suppressed because it is too large Load Diff
modules/dnn/test/test_int8_layers.cpp
+14 -97
View File
@@ -648,9 +648,9 @@ public:
normAssert(ref, out, "", l1, lInf);
}
void testDarknetModel(const std::string& cfg, const std::string& weights,
const cv::Mat& ref, double scoreDiff, double iouDiff,
float confThreshold = 0.24, float nmsThreshold = 0.4, bool perChannel = true)
void testYOLOModel(const std::string& model,
const cv::Mat& ref, double scoreDiff, double iouDiff,
float confThreshold = 0.24, float nmsThreshold = 0.4, bool perChannel = true)
{
CV_Assert(ref.cols == 7);
std::vector<std::vector<int> > refClassIds;
@@ -689,7 +689,7 @@ public:
Mat inp = blobFromImages(samples, 1.0/255, Size(416, 416), Scalar(), true, false);
Net baseNet = readNetFromDarknet(findDataFile("dnn/" + cfg), findDataFile("dnn/" + weights, false));
Net baseNet = readNet(findDataFile("dnn/" + model, false));
Net qnet = baseNet.quantize(inp, CV_32F, CV_32F, perChannel);
qnet.setPreferableBackend(backend);
qnet.setPreferableTarget(target);
@@ -1225,86 +1225,6 @@ TEST_P(Test_Int8_nets, RFCN)
testFaster(net, ref, confThreshold, scoreDiff, iouDiff);
}
TEST_P(Test_Int8_nets, YoloVoc)
{
applyTestTag(
#if defined(OPENCV_32BIT_CONFIGURATION) && defined(HAVE_OPENCL)
CV_TEST_TAG_MEMORY_2GB,
#else
CV_TEST_TAG_MEMORY_1GB,
#endif
CV_TEST_TAG_LONG,
CV_TEST_TAG_DEBUG_VERYLONG
);
if (target == DNN_TARGET_OPENCL_FP16 && !ocl::Device::getDefault().isIntel())
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
if (target == DNN_TARGET_OPENCL && !ocl::Device::getDefault().isIntel())
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL);
Mat ref = (Mat_<float>(6, 7) << 0, 6, 0.750469f, 0.577374f, 0.127391f, 0.902949f, 0.300809f,
0, 1, 0.780879f, 0.270762f, 0.264102f, 0.732475f, 0.745412f,
0, 11, 0.901615f, 0.1386f, 0.338509f, 0.421337f, 0.938789f,
1, 14, 0.623813f, 0.183179f, 0.381921f, 0.247726f, 0.625847f,
1, 6, 0.667770f, 0.446555f, 0.453578f, 0.499986f, 0.519167f,
1, 6, 0.844947f, 0.637058f, 0.460398f, 0.828508f, 0.66427f);
std::string config_file = "yolo-voc.cfg";
std::string weights_file = "yolo-voc.weights";
double scoreDiff = 0.12, iouDiff = 0.3;
{
SCOPED_TRACE("batch size 1");
testDarknetModel(config_file, weights_file, ref.rowRange(0, 3), scoreDiff, iouDiff);
}
{
SCOPED_TRACE("batch size 2");
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff);
}
}
TEST_P(Test_Int8_nets, TinyYoloVoc)
{
applyTestTag(
CV_TEST_TAG_MEMORY_512MB,
CV_TEST_TAG_DEBUG_VERYLONG
);
if (target == DNN_TARGET_OPENCL_FP16 && !ocl::Device::getDefault().isIntel())
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
if (target == DNN_TARGET_OPENCL && !ocl::Device::getDefault().isIntel())
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL);
Mat ref = (Mat_<float>(4, 7) << 0, 6, 0.761967f, 0.579042f, 0.159161f, 0.894482f, 0.31994f,
0, 11, 0.780595f, 0.129696f, 0.386467f, 0.445275f, 0.920994f,
1, 6, 0.651450f, 0.460526f, 0.458019f, 0.522527f, 0.5341f,
1, 6, 0.928758f, 0.651024f, 0.463539f, 0.823784f, 0.654998f);
std::string config_file = "tiny-yolo-voc.cfg";
std::string weights_file = "tiny-yolo-voc.weights";
double scoreDiff = 0.043, iouDiff = 0.12;
{
SCOPED_TRACE("batch size 1");
testDarknetModel(config_file, weights_file, ref.rowRange(0, 2), scoreDiff, iouDiff);
{
SCOPED_TRACE("Per-tensor quantize");
testDarknetModel(config_file, weights_file, ref.rowRange(0, 2), 0.1, 0.2, 0.24, 0.6, false);
}
}
{
SCOPED_TRACE("batch size 2");
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff);
{
SCOPED_TRACE("Per-tensor quantize");
testDarknetModel(config_file, weights_file, ref, 0.1, 0.2, 0.24, 0.6, false);
}
}
}
TEST_P(Test_Int8_nets, YOLOv3)
{
applyTestTag(
@@ -1334,18 +1254,17 @@ TEST_P(Test_Int8_nets, YOLOv3)
};
Mat ref(N0 + N1, 7, CV_32FC1, (void*)ref_);
std::string config_file = "yolov3.cfg";
std::string weights_file = "yolov3.weights";
std::string model_file = "yolov3.onnx";
double scoreDiff = 0.08, iouDiff = 0.21, confThreshold = 0.25;
{
SCOPED_TRACE("batch size 1");
testDarknetModel(config_file, weights_file, ref.rowRange(0, N0), scoreDiff, iouDiff, confThreshold);
testYOLOModel(model_file, ref.rowRange(0, N0), scoreDiff, iouDiff, confThreshold);
}
{
SCOPED_TRACE("batch size 2");
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff, confThreshold);
testYOLOModel(model_file, ref, scoreDiff, iouDiff, confThreshold);
}
}
@@ -1379,18 +1298,17 @@ TEST_P(Test_Int8_nets, YOLOv4)
};
Mat ref(N0 + N1, 7, CV_32FC1, (void*)ref_);
std::string config_file = "yolov4.cfg";
std::string weights_file = "yolov4.weights";
std::string model_file = "yolov4.onnx";
double scoreDiff = 0.15, iouDiff = 0.2;
{
SCOPED_TRACE("batch size 1");
testDarknetModel(config_file, weights_file, ref.rowRange(0, N0), scoreDiff, iouDiff);
testYOLOModel(model_file, ref.rowRange(0, N0), scoreDiff, iouDiff);
}
{
SCOPED_TRACE("batch size 2");
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff);
testYOLOModel(model_file, ref, scoreDiff, iouDiff);
}
}
@@ -1421,18 +1339,17 @@ TEST_P(Test_Int8_nets, YOLOv4_tiny)
};
Mat ref(N0 + N1, 7, CV_32FC1, (void*)ref_);
std::string config_file = "yolov4-tiny-2020-12.cfg";
std::string weights_file = "yolov4-tiny-2020-12.weights";
std::string model_file = "yolov4-tiny.onnx";
double scoreDiff = 0.12;
double iouDiff = target == DNN_TARGET_OPENCL_FP16 ? 0.2 : 0.118;
{
SCOPED_TRACE("batch size 1");
testDarknetModel(config_file, weights_file, ref.rowRange(0, N0), scoreDiff, iouDiff, confThreshold);
testYOLOModel(model_file, ref.rowRange(0, N0), scoreDiff, iouDiff, confThreshold);
{
SCOPED_TRACE("Per-tensor quantize");
testDarknetModel(config_file, weights_file, ref.rowRange(0, N0), scoreDiff, 0.224, 0.7, 0.4, false);
testYOLOModel(model_file, ref.rowRange(0, N0), scoreDiff, 0.224, 0.7, 0.4, false);
}
}
@@ -1440,7 +1357,7 @@ TEST_P(Test_Int8_nets, YOLOv4_tiny)
/* bad accuracy on second image
{
SCOPED_TRACE("batch size 2");
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff, confThreshold);
testYOLOModel(model_file, ref, scoreDiff, iouDiff, confThreshold);
}
*/
}
modules/dnn/test/test_misc.cpp
-3
View File
@@ -247,9 +247,6 @@ TEST(readNet, Regression)
Net net = readNet(findDataFile("dnn/squeezenet_v1.1.prototxt"),
findDataFile("dnn/squeezenet_v1.1.caffemodel", false));
EXPECT_FALSE(net.empty());
net = readNet(findDataFile("dnn/tiny-yolo-voc.cfg"),
findDataFile("dnn/tiny-yolo-voc.weights", false));
EXPECT_FALSE(net.empty());
net = readNet(findDataFile("dnn/ssd_mobilenet_v1_coco.pbtxt"),
findDataFile("dnn/ssd_mobilenet_v1_coco.pb", false));
EXPECT_FALSE(net.empty());
modules/dnn/test/test_model.cpp
-123
View File
@@ -289,129 +289,6 @@ TEST_P(Test_Model, Classify)
}
TEST_P(Test_Model, DetectRegion)
{
applyTestTag(
CV_TEST_TAG_MEMORY_2GB,
CV_TEST_TAG_LONG,
CV_TEST_TAG_DEBUG_VERYLONG
);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2022010000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#elif defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#elif defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2020040000) // nGraph compilation failure
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#elif defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019010000)
// FIXIT DNN_BACKEND_INFERENCE_ENGINE is misused
if (backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);
#endif
#if defined(INF_ENGINE_RELEASE)
if (target == DNN_TARGET_MYRIAD
&& getInferenceEngineVPUType() == CV_DNN_INFERENCE_ENGINE_VPU_TYPE_MYRIAD_X)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD_X);
#endif
std::vector<int> refClassIds = {6, 1, 11};
std::vector<float> refConfidences = {0.750469f, 0.780879f, 0.901615f};
std::vector<Rect2d> refBoxes = {Rect2d(240, 53, 135, 72),
Rect2d(112, 109, 192, 200),
Rect2d(58, 141, 117, 249)};
std::string img_path = _tf("dog416.png");
std::string weights_file = _tf("yolo-voc.weights", false);
std::string config_file = _tf("yolo-voc.cfg");
double scale = 1.0 / 255.0;
Size size{416, 416};
bool swapRB = true;
double confThreshold = 0.24;
double nmsThreshold = (target == DNN_TARGET_MYRIAD) ? 0.397 : 0.4;
double scoreDiff = 8e-5, iouDiff = 1e-5;
if (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD || target == DNN_TARGET_CUDA_FP16 || target == DNN_TARGET_CPU_FP16)
{
scoreDiff = 1e-2;
iouDiff = 1.6e-2;
}
testDetectModel(weights_file, config_file, img_path, refClassIds, refConfidences,
refBoxes, scoreDiff, iouDiff, confThreshold, nmsThreshold, size,
Scalar(), scale, swapRB);
}
TEST_P(Test_Model, DetectRegionWithNmsAcrossClasses)
{
applyTestTag(
CV_TEST_TAG_MEMORY_2GB,
CV_TEST_TAG_LONG,
CV_TEST_TAG_DEBUG_VERYLONG
);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2022010000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#elif defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#elif defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2020040000) // nGraph compilation failure
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#elif defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019010000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);
#endif
#if defined(INF_ENGINE_RELEASE)
if (target == DNN_TARGET_MYRIAD
&& getInferenceEngineVPUType() == CV_DNN_INFERENCE_ENGINE_VPU_TYPE_MYRIAD_X)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD_X);
#endif
std::vector<int> refClassIds = { 6, 11 };
std::vector<float> refConfidences = { 0.750469f, 0.901615f };
std::vector<Rect2d> refBoxes = { Rect2d(240, 53, 135, 72),
Rect2d(58, 141, 117, 249) };
std::string img_path = _tf("dog416.png");
std::string weights_file = _tf("yolo-voc.weights", false);
std::string config_file = _tf("yolo-voc.cfg");
double scale = 1.0 / 255.0;
Size size{ 416, 416 };
bool swapRB = true;
bool crop = false;
bool nmsAcrossClasses = true;
double confThreshold = 0.24;
double nmsThreshold = (target == DNN_TARGET_MYRIAD) ? 0.15: 0.15;
double scoreDiff = 8e-5, iouDiff = 1e-5;
if (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD || target == DNN_TARGET_CUDA_FP16 || target == DNN_TARGET_CPU_FP16)
{
scoreDiff = 1e-2;
iouDiff = 1.6e-2;
}
testDetectModel(weights_file, config_file, img_path, refClassIds, refConfidences,
refBoxes, scoreDiff, iouDiff, confThreshold, nmsThreshold, size,
Scalar(), scale, swapRB, crop,
nmsAcrossClasses);
}
TEST_P(Test_Model, DetectionOutput)
{
applyTestTag(CV_TEST_TAG_DEBUG_VERYLONG);