mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2024-11-29 13:47:32 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #10164 from pengli:dnn

Browse Source
This commit is contained in:
Alexander Alekhin 2017-11-29 12:05:10 +00:00
commit cc2ee923e4
4 changed files with 219 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
modules/dnn/src/dnn.cpp
View File

@ -1196,7 +1196,8 @@ struct Net::Impl
// some other layers.
// TODO: OpenCL target support more fusion styles.
if ( preferableTarget == DNN_TARGET_OPENCL && ld.layerInstance->type.compare("Convolution") )
if ( preferableTarget == DNN_TARGET_OPENCL &&
(!cv::ocl::useOpenCL() || ld.layerInstance->type.compare("Convolution")) )
continue;
Ptr<Layer>& currLayer = ld.layerInstance;
@ -1214,7 +1215,10 @@ struct Net::Impl
{
printf_(("\tfused with %s\n", nextBNormLayer->name.c_str()));
bnormData->skipFlags[DNN_BACKEND_DEFAULT] = true;
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
if ( preferableTarget == DNN_TARGET_OPENCL )
ld.umat_outputBlobs = layers[lpNext.lid].umat_outputBlobs;
else
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
if( bnormData->consumers.size() == 1 )
{
nextData = &layers[bnormData->consumers[0].lid];
@ -1234,7 +1238,10 @@ struct Net::Impl
{
printf_(("\tfused with %s\n", nextScaleLayer->name.c_str()));
scaleData->skipFlags[DNN_BACKEND_DEFAULT] = true;
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
if ( preferableTarget == DNN_TARGET_OPENCL )
ld.umat_outputBlobs = layers[lpNext.lid].umat_outputBlobs;
else
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
if( scaleData->consumers.size() == 1 )
{
nextData = &layers[scaleData->consumers[0].lid];
@ -1263,7 +1270,10 @@ struct Net::Impl
LayerData *activData = nextData;
printf_(("\tfused with %s\n", nextActivLayer->name.c_str()));
activData->skipFlags[DNN_BACKEND_DEFAULT] = true;
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
if ( preferableTarget == DNN_TARGET_OPENCL )
ld.umat_outputBlobs = layers[lpNext.lid].umat_outputBlobs;
else
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
if ( preferableTarget == DNN_TARGET_OPENCL )
{
@ -1325,13 +1335,13 @@ struct Net::Impl
!nextData->type.compare("Power")) &&
currLayer->setActivation(nextActivLayer) )
{
CV_Assert(firstConvLayerData->outputBlobs.size() == 1 && ld.inputBlobs.size() == 1);
ld.inputBlobs.push_back(&firstConvLayerData->outputBlobs[0]);
CV_Assert(firstConvLayerData->umat_outputBlobs.size() == 1 && ld.umat_inputBlobs.size() == 1);
ld.umat_inputBlobs.push_back(firstConvLayerData->umat_outputBlobs[0]);
printf_(("\tfused with %s\n", nextEltwiseLayer->name.c_str()));
printf_(("\tfused with %s\n", nextActivLayer->name.c_str()));
eltwiseData->skipFlags[DNN_BACKEND_DEFAULT] = true;
nextData->skipFlags[DNN_BACKEND_DEFAULT] = true;
ld.outputBlobs = layers[lpNext.lid].outputBlobs;
ld.umat_outputBlobs = layers[lpNext.lid].umat_outputBlobs;
}
}
}

107
modules/dnn/test/test_caffe_importer.cpp
View File

@ -42,6 +42,8 @@
#include "test_precomp.hpp"
#include "npy_blob.hpp"
#include <opencv2/dnn/shape_utils.hpp>
#include <opencv2/core/ocl.hpp>
#include <opencv2/ts/ocl_test.hpp>
namespace cvtest
{
@ -119,6 +121,43 @@ TEST_P(Reproducibility_AlexNet, Accuracy)
INSTANTIATE_TEST_CASE_P(Test_Caffe, Reproducibility_AlexNet, testing::Values(true, false));
typedef testing::TestWithParam<tuple<bool> > Reproducibility_OCL_AlexNet;
OCL_TEST_P(Reproducibility_OCL_AlexNet, Accuracy)
{
bool readFromMemory = get<0>(GetParam());
Net net;
{
const string proto = findDataFile("dnn/bvlc_alexnet.prototxt", false);
const string model = findDataFile("dnn/bvlc_alexnet.caffemodel", false);
if (readFromMemory)
{
string dataProto;
ASSERT_TRUE(readFileInMemory(proto, dataProto));
string dataModel;
ASSERT_TRUE(readFileInMemory(model, dataModel));
net = readNetFromCaffe(dataProto.c_str(), dataProto.size(),
dataModel.c_str(), dataModel.size());
}
else
net = readNetFromCaffe(proto, model);
ASSERT_FALSE(net.empty());
}
net.setPreferableBackend(DNN_BACKEND_DEFAULT);
net.setPreferableTarget(DNN_TARGET_OPENCL);
Mat sample = imread(_tf("grace_hopper_227.png"));
ASSERT_TRUE(!sample.empty());
net.setInput(blobFromImage(sample, 1.0f, Size(227, 227), Scalar(), false), "data");
Mat out = net.forward("prob");
Mat ref = blobFromNPY(_tf("caffe_alexnet_prob.npy"));
normAssert(ref, out);
}
OCL_INSTANTIATE_TEST_CASE_P(Test_Caffe, Reproducibility_OCL_AlexNet, testing::Values(true, false));
#if !defined(_WIN32) || defined(_WIN64)
TEST(Reproducibility_FCN, Accuracy)
{
@ -201,6 +240,38 @@ TEST(Reproducibility_MobileNet_SSD, Accuracy)
}
}
OCL_TEST(Reproducibility_MobileNet_SSD, Accuracy)
{
const string proto = findDataFile("dnn/MobileNetSSD_deploy.prototxt", false);
const string model = findDataFile("dnn/MobileNetSSD_deploy.caffemodel", false);
Net net = readNetFromCaffe(proto, model);
net.setPreferableBackend(DNN_BACKEND_DEFAULT);
net.setPreferableTarget(DNN_TARGET_OPENCL);
Mat sample = imread(_tf("street.png"));
Mat inp = blobFromImage(sample, 1.0f / 127.5, Size(300, 300), Scalar(127.5, 127.5, 127.5), false);
net.setInput(inp);
Mat out = net.forward();
Mat ref = blobFromNPY(_tf("mobilenet_ssd_caffe_out.npy"));
normAssert(ref, out);
// Check that detections aren't preserved.
inp.setTo(0.0f);
net.setInput(inp);
out = net.forward();
const int numDetections = out.size[2];
ASSERT_NE(numDetections, 0);
for (int i = 0; i < numDetections; ++i)
{
float confidence = out.ptr<float>(0, 0, i)[2];
ASSERT_EQ(confidence, 0);
}
}
TEST(Reproducibility_ResNet50, Accuracy)
{
Net net = readNetFromCaffe(findDataFile("dnn/ResNet-50-deploy.prototxt", false),
@ -216,6 +287,24 @@ TEST(Reproducibility_ResNet50, Accuracy)
normAssert(ref, out);
}
OCL_TEST(Reproducibility_ResNet50, Accuracy)
{
Net net = readNetFromCaffe(findDataFile("dnn/ResNet-50-deploy.prototxt", false),
findDataFile("dnn/ResNet-50-model.caffemodel", false));
net.setPreferableBackend(DNN_BACKEND_DEFAULT);
net.setPreferableTarget(DNN_TARGET_OPENCL);
Mat input = blobFromImage(imread(_tf("googlenet_0.png")), 1.0f, Size(224,224), Scalar(), false);
ASSERT_TRUE(!input.empty());
net.setInput(input);
Mat out = net.forward();
Mat ref = blobFromNPY(_tf("resnet50_prob.npy"));
normAssert(ref, out);
}
TEST(Reproducibility_SqueezeNet_v1_1, Accuracy)
{
Net net = readNetFromCaffe(findDataFile("dnn/squeezenet_v1.1.prototxt", false),
@ -231,6 +320,24 @@ TEST(Reproducibility_SqueezeNet_v1_1, Accuracy)
normAssert(ref, out);
}
OCL_TEST(Reproducibility_SqueezeNet_v1_1, Accuracy)
{
Net net = readNetFromCaffe(findDataFile("dnn/squeezenet_v1.1.prototxt", false),
findDataFile("dnn/squeezenet_v1.1.caffemodel", false));
net.setPreferableBackend(DNN_BACKEND_DEFAULT);
net.setPreferableTarget(DNN_TARGET_OPENCL);
Mat input = blobFromImage(imread(_tf("googlenet_0.png")), 1.0f, Size(227,227), Scalar(), false);
ASSERT_TRUE(!input.empty());
net.setInput(input);
Mat out = net.forward();
Mat ref = blobFromNPY(_tf("squeezenet_v1.1_prob.npy"));
normAssert(ref, out);
}
TEST(Reproducibility_AlexNet_fp16, Accuracy)
{
const float l1 = 1e-5;

62
modules/dnn/test/test_darknet_importer.cpp
View File

@ -184,6 +184,68 @@ TEST(Reproducibility_TinyYoloVoc, Accuracy)
normAssert(ref, detection);
}
OCL_TEST(Reproducibility_YoloVoc, Accuracy)
{
Net net;
{
const string cfg = findDataFile("dnn/yolo-voc.cfg", false);
const string model = findDataFile("dnn/yolo-voc.weights", false);
net = readNetFromDarknet(cfg, model);
ASSERT_FALSE(net.empty());
}
net.setPreferableBackend(DNN_BACKEND_DEFAULT);
net.setPreferableTarget(DNN_TARGET_OPENCL);
// dog416.png is dog.jpg that resized to 416x416 in the lossless PNG format
Mat sample = imread(_tf("dog416.png"));
ASSERT_TRUE(!sample.empty());
Size inputSize(416, 416);
if (sample.size() != inputSize)
resize(sample, sample, inputSize);
net.setInput(blobFromImage(sample, 1 / 255.F), "data");
Mat out = net.forward("detection_out");
Mat detection;
const float confidenceThreshold = 0.24;
for (int i = 0; i < out.rows; i++) {
const int probability_index = 5;
const int probability_size = out.cols - probability_index;
float *prob_array_ptr = &out.at<float>(i, probability_index);
size_t objectClass = std::max_element(prob_array_ptr, prob_array_ptr + probability_size) - prob_array_ptr;
float confidence = out.at<float>(i, (int)objectClass + probability_index);
if (confidence > confidenceThreshold)
detection.push_back(out.row(i));
}
// obtained by: ./darknet detector test ./cfg/voc.data ./cfg/yolo-voc.cfg ./yolo-voc.weights -thresh 0.24 ./dog416.png
// There are 3 objects (6-car, 1-bicycle, 11-dog) with 25 values for each:
// { relative_center_x, relative_center_y, relative_width, relative_height, unused_t0, probability_for_each_class[20] }
float ref_array[] = {
0.740161F, 0.214100F, 0.325575F, 0.173418F, 0.750769F, 0.000000F, 0.000000F, 0.000000F, 0.000000F,
0.000000F, 0.000000F, 0.750469F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F,
0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F,
0.501618F, 0.504757F, 0.461713F, 0.481310F, 0.783550F, 0.000000F, 0.780879F, 0.000000F, 0.000000F,
0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F,
0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F,
0.279968F, 0.638651F, 0.282737F, 0.600284F, 0.901864F, 0.000000F, 0.000000F, 0.000000F, 0.000000F,
0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.901615F,
0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F, 0.000000F
};
const int number_of_objects = 3;
Mat ref(number_of_objects, sizeof(ref_array) / (number_of_objects * sizeof(float)), CV_32FC1, &ref_array);
normAssert(ref, detection);
}
TEST(Reproducibility_YoloVoc, Accuracy)
{
Net net;

33
modules/dnn/test/test_torch_importer.cpp
View File

@ -382,6 +382,39 @@ TEST(Torch_Importer, FastNeuralStyle_accuracy)
}
}
OCL_TEST(Torch_Importer, FastNeuralStyle_accuracy)
{
std::string models[] = {"dnn/fast_neural_style_eccv16_starry_night.t7",
"dnn/fast_neural_style_instance_norm_feathers.t7"};
std::string targets[] = {"dnn/lena_starry_night.png", "dnn/lena_feathers.png"};
for (int i = 0; i < 2; ++i)
{
const string model = findDataFile(models[i], false);
Net net = readNetFromTorch(model);
net.setPreferableBackend(DNN_BACKEND_DEFAULT);
net.setPreferableTarget(DNN_TARGET_OPENCL);
Mat img = imread(findDataFile("dnn/googlenet_1.png", false));
Mat inputBlob = blobFromImage(img, 1.0, Size(), Scalar(103.939, 116.779, 123.68), false);
net.setInput(inputBlob);
Mat out = net.forward();
// Deprocessing.
getPlane(out, 0, 0) += 103.939;
getPlane(out, 0, 1) += 116.779;
getPlane(out, 0, 2) += 123.68;
out = cv::min(cv::max(0, out), 255);
Mat ref = imread(findDataFile(targets[i]));
Mat refBlob = blobFromImage(ref, 1.0, Size(), Scalar(), false);
normAssert(out, refBlob, "", 0.5, 1.1);
}
}
}
#endif