potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-30 07:43:03 +04:00
Code Packages Releases Wiki Activity

Faster-RCNN anf RFCN models on CPU using Intel's Inference Engine backend.

Browse Source 
Enable Torch layers tests with Intel's Inference Engine backend.
This commit is contained in:
Dmitry Kurtaev
2018-07-24 19:12:58 +03:00
parent 236f383969
commit faa6c4e1e1
12 changed files with 309 additions and 159 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/test/test_caffe_importer.cpp
+70 -35
View File
@@ -51,6 +51,33 @@ static std::string _tf(TString filename)
return (getOpenCVExtraDir() + "/dnn/") + filename;
}
class Test_Caffe_nets : public DNNTestLayer
{
public:
void testFaster(const std::string& proto, const std::string& model, const Mat& ref,
double scoreDiff = 0.0, double iouDiff = 0.0)
{
checkBackend();
Net net = readNetFromCaffe(findDataFile("dnn/" + proto, false),
findDataFile("dnn/" + model, false));
net.setPreferableBackend(backend);
net.setPreferableTarget(target);
Mat img = imread(findDataFile("dnn/dog416.png", false));
resize(img, img, Size(800, 600));
Mat blob = blobFromImage(img, 1.0, Size(), Scalar(102.9801, 115.9465, 122.7717), false, false);
Mat imInfo = (Mat_<float>(1, 3) << img.rows, img.cols, 1.6f);
net.setInput(blob, "data");
net.setInput(imInfo, "im_info");
// Output has shape 1x1xNx7 where N - number of detections.
// An every detection is a vector of values [id, classId, confidence, left, top, right, bottom]
Mat out = net.forward();
scoreDiff = scoreDiff ? scoreDiff : default_l1;
iouDiff = iouDiff ? iouDiff : default_lInf;
normAssertDetections(ref, out, ("model name: " + model).c_str(), 0.8, scoreDiff, iouDiff);
}
};
TEST(Test_Caffe, memory_read)
{
const string proto = findDataFile("dnn/bvlc_googlenet.prototxt", false);
@@ -344,9 +371,15 @@ TEST(Reproducibility_GoogLeNet_fp16, Accuracy)
}
// https://github.com/richzhang/colorization
TEST(Reproducibility_Colorization, Accuracy)
TEST_P(Test_Caffe_nets, Colorization)
{
const float l1 = 3e-5;
checkBackend();
if ((backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_OPENCL_FP16) ||
(backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_MYRIAD) ||
(backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16))
throw SkipTestException("");
const float l1 = 4e-4;
const float lInf = 3e-3;
Mat inp = blobFromNPY(_tf("colorization_inp.npy"));
@@ -356,7 +389,8 @@ TEST(Reproducibility_Colorization, Accuracy)
const string proto = findDataFile("dnn/colorization_deploy_v2.prototxt", false);
const string model = findDataFile("dnn/colorization_release_v2.caffemodel", false);
Net net = readNetFromCaffe(proto, model);
net.setPreferableBackend(DNN_BACKEND_OPENCV);
net.setPreferableBackend(backend);
net.setPreferableTarget(target);
net.getLayer(net.getLayerId("class8_ab"))->blobs.push_back(kernel);
net.getLayer(net.getLayerId("conv8_313_rh"))->blobs.push_back(Mat(1, 313, CV_32F, 2.606));
@@ -447,39 +481,40 @@ INSTANTIATE_TEST_CASE_P(Test_Caffe, opencv_face_detector,
)
);
TEST(Test_Caffe, FasterRCNN_and_RFCN)
TEST_P(Test_Caffe_nets, FasterRCNN_vgg16)
{
std::string models[] = {"VGG16_faster_rcnn_final.caffemodel", "ZF_faster_rcnn_final.caffemodel",
"resnet50_rfcn_final.caffemodel"};
std::string protos[] = {"faster_rcnn_vgg16.prototxt", "faster_rcnn_zf.prototxt",
"rfcn_pascal_voc_resnet50.prototxt"};
Mat refs[] = {(Mat_<float>(3, 7) << 0, 2, 0.949398, 99.2454, 210.141, 601.205, 462.849,
0, 7, 0.997022, 481.841, 92.3218, 722.685, 175.953,
0, 12, 0.993028, 133.221, 189.377, 350.994, 563.166),
(Mat_<float>(3, 7) << 0, 2, 0.90121, 120.407, 115.83, 570.586, 528.395,
0, 7, 0.988779, 469.849, 75.1756, 718.64, 186.762,
0, 12, 0.967198, 138.588, 206.843, 329.766, 553.176),
(Mat_<float>(2, 7) << 0, 7, 0.991359, 491.822, 81.1668, 702.573, 178.234,
0, 12, 0.94786, 132.093, 223.903, 338.077, 566.16)};
for (int i = 0; i < 3; ++i)
{
std::string proto = findDataFile("dnn/" + protos[i], false);
std::string model = findDataFile("dnn/" + models[i], false);
Net net = readNetFromCaffe(proto, model);
net.setPreferableBackend(DNN_BACKEND_OPENCV);
Mat img = imread(findDataFile("dnn/dog416.png", false));
resize(img, img, Size(800, 600));
Mat blob = blobFromImage(img, 1.0, Size(), Scalar(102.9801, 115.9465, 122.7717), false, false);
Mat imInfo = (Mat_<float>(1, 3) << img.rows, img.cols, 1.6f);
net.setInput(blob, "data");
net.setInput(imInfo, "im_info");
// Output has shape 1x1xNx7 where N - number of detections.
// An every detection is a vector of values [id, classId, confidence, left, top, right, bottom]
Mat out = net.forward();
normAssertDetections(refs[i], out, ("model name: " + models[i]).c_str(), 0.8);
}
if ((backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_MYRIAD) ||
(backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16))
throw SkipTestException("");
static Mat ref = (Mat_<float>(3, 7) << 0, 2, 0.949398, 99.2454, 210.141, 601.205, 462.849,
0, 7, 0.997022, 481.841, 92.3218, 722.685, 175.953,
0, 12, 0.993028, 133.221, 189.377, 350.994, 563.166);
testFaster("faster_rcnn_vgg16.prototxt", "VGG16_faster_rcnn_final.caffemodel", ref);
}
TEST_P(Test_Caffe_nets, FasterRCNN_zf)
{
if ((backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_OPENCL_FP16) ||
(backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_MYRIAD) ||
(backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16))
throw SkipTestException("");
static Mat ref = (Mat_<float>(3, 7) << 0, 2, 0.90121, 120.407, 115.83, 570.586, 528.395,
0, 7, 0.988779, 469.849, 75.1756, 718.64, 186.762,
0, 12, 0.967198, 138.588, 206.843, 329.766, 553.176);
testFaster("faster_rcnn_zf.prototxt", "ZF_faster_rcnn_final.caffemodel", ref);
}
TEST_P(Test_Caffe_nets, RFCN)
{
if ((backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_OPENCL_FP16) ||
(backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_MYRIAD) ||
(backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16))
throw SkipTestException("");
static Mat ref = (Mat_<float>(2, 7) << 0, 7, 0.991359, 491.822, 81.1668, 702.573, 178.234,
0, 12, 0.94786, 132.093, 223.903, 338.077, 566.16);
testFaster("rfcn_pascal_voc_resnet50.prototxt", "resnet50_rfcn_final.caffemodel", ref);
}
INSTANTIATE_TEST_CASE_P(/**/, Test_Caffe_nets, dnnBackendsAndTargets());
}} // namespace
modules/dnn/test/test_layers.cpp
+1 -9
View File
@@ -1205,14 +1205,6 @@ public:
}
}
void forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs, OutputArrayOfArrays internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs, outputs, internals);
}
private:
int outWidth, outHeight, zoomFactor;
};
@@ -1225,7 +1217,7 @@ TEST_P(Test_Caffe_layers, DISABLED_Interp) // requires patched protobuf (availa
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_MYRIAD)
throw SkipTestException("");
// Test a cusom layer.
// Test a custom layer.
CV_DNN_REGISTER_LAYER_CLASS(Interp, CustomInterpLayer);
try
{
modules/dnn/test/test_torch_importer.cpp
+110 -79
View File
@@ -69,100 +69,119 @@ TEST(Torch_Importer, simple_read)
ASSERT_FALSE(net.empty());
}
static void runTorchNet(String prefix, int targetId = DNN_TARGET_CPU, String outLayerName = "",
bool check2ndBlob = false, bool isBinary = false)
class Test_Torch_layers : public DNNTestLayer
{
String suffix = (isBinary) ? ".dat" : ".txt";
Net net = readNetFromTorch(_tf(prefix + "_net" + suffix), isBinary);
ASSERT_FALSE(net.empty());
net.setPreferableBackend(DNN_BACKEND_OPENCV);
net.setPreferableTarget(targetId);
Mat inp, outRef;
ASSERT_NO_THROW( inp = readTorchBlob(_tf(prefix + "_input" + suffix), isBinary) );
ASSERT_NO_THROW( outRef = readTorchBlob(_tf(prefix + "_output" + suffix), isBinary) );
if (outLayerName.empty())
outLayerName = net.getLayerNames().back();
net.setInput(inp);
std::vector<Mat> outBlobs;
net.forward(outBlobs, outLayerName);
normAssert(outRef, outBlobs[0]);
if (check2ndBlob)
public:
void runTorchNet(const String& prefix, String outLayerName = "",
bool check2ndBlob = false, bool isBinary = false,
double l1 = 0.0, double lInf = 0.0)
{
Mat out2 = outBlobs[1];
Mat ref2 = readTorchBlob(_tf(prefix + "_output_2" + suffix), isBinary);
normAssert(out2, ref2);
}
}
String suffix = (isBinary) ? ".dat" : ".txt";
typedef testing::TestWithParam<Target> Test_Torch_layers;
Mat inp, outRef;
ASSERT_NO_THROW( inp = readTorchBlob(_tf(prefix + "_input" + suffix), isBinary) );
ASSERT_NO_THROW( outRef = readTorchBlob(_tf(prefix + "_output" + suffix), isBinary) );
checkBackend(backend, target, &inp, &outRef);
Net net = readNetFromTorch(_tf(prefix + "_net" + suffix), isBinary);
ASSERT_FALSE(net.empty());
net.setPreferableBackend(backend);
net.setPreferableTarget(target);
if (outLayerName.empty())
outLayerName = net.getLayerNames().back();
net.setInput(inp);
std::vector<Mat> outBlobs;
net.forward(outBlobs, outLayerName);
l1 = l1 ? l1 : default_l1;
lInf = lInf ? lInf : default_lInf;
normAssert(outRef, outBlobs[0], "", l1, lInf);
if (check2ndBlob && backend != DNN_BACKEND_INFERENCE_ENGINE)
{
Mat out2 = outBlobs[1];
Mat ref2 = readTorchBlob(_tf(prefix + "_output_2" + suffix), isBinary);
normAssert(out2, ref2, "", l1, lInf);
}
}
};
TEST_P(Test_Torch_layers, run_convolution)
{
runTorchNet("net_conv", GetParam(), "", false, true);
if ((backend == DNN_BACKEND_INFERENCE_ENGINE && target != DNN_TARGET_CPU) ||
(backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16))
throw SkipTestException("");
runTorchNet("net_conv", "", false, true);
}
TEST_P(Test_Torch_layers, run_pool_max)
{
runTorchNet("net_pool_max", GetParam(), "", true);
if (backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16)
throw SkipTestException("");
runTorchNet("net_pool_max", "", true);
}
TEST_P(Test_Torch_layers, run_pool_ave)
{
runTorchNet("net_pool_ave", GetParam());
runTorchNet("net_pool_ave");
}
TEST_P(Test_Torch_layers, run_reshape)
{
int targetId = GetParam();
runTorchNet("net_reshape", targetId);
runTorchNet("net_reshape_batch", targetId);
runTorchNet("net_reshape_single_sample", targetId);
runTorchNet("net_reshape_channels", targetId, "", false, true);
runTorchNet("net_reshape");
runTorchNet("net_reshape_batch");
runTorchNet("net_reshape_channels", "", false, true);
}
TEST_P(Test_Torch_layers, run_reshape_single_sample)
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_OPENCL_FP16)
throw SkipTestException("");
runTorchNet("net_reshape_single_sample", "", false, false,
(target == DNN_TARGET_MYRIAD || target == DNN_TARGET_OPENCL_FP16) ? 0.0052 : 0.0);
}
TEST_P(Test_Torch_layers, run_linear)
{
runTorchNet("net_linear_2d", GetParam());
if (backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16)
throw SkipTestException("");
runTorchNet("net_linear_2d");
}
TEST_P(Test_Torch_layers, run_concat)
{
int targetId = GetParam();
runTorchNet("net_concat", targetId, "l5_torchMerge");
runTorchNet("net_depth_concat", targetId, "", false, true);
runTorchNet("net_concat", "l5_torchMerge");
runTorchNet("net_depth_concat", "", false, true, 0.0,
target == DNN_TARGET_OPENCL_FP16 ? 0.021 : 0.0);
}
TEST_P(Test_Torch_layers, run_deconv)
{
runTorchNet("net_deconv", GetParam());
runTorchNet("net_deconv");
}
TEST_P(Test_Torch_layers, run_batch_norm)
{
runTorchNet("net_batch_norm", GetParam(), "", false, true);
runTorchNet("net_batch_norm", "", false, true);
}
TEST_P(Test_Torch_layers, net_prelu)
{
runTorchNet("net_prelu", GetParam());
runTorchNet("net_prelu");
}
TEST_P(Test_Torch_layers, net_cadd_table)
{
runTorchNet("net_cadd_table", GetParam());
runTorchNet("net_cadd_table");
}
TEST_P(Test_Torch_layers, net_softmax)
{
int targetId = GetParam();
runTorchNet("net_softmax", targetId);
runTorchNet("net_softmax_spatial", targetId);
runTorchNet("net_softmax");
runTorchNet("net_softmax_spatial");
}
TEST_P(Test_Torch_layers, net_logsoftmax)
@@ -173,40 +192,55 @@ TEST_P(Test_Torch_layers, net_logsoftmax)
TEST_P(Test_Torch_layers, net_lp_pooling)
{
int targetId = GetParam();
runTorchNet("net_lp_pooling_square", targetId, "", false, true);
runTorchNet("net_lp_pooling_power", targetId, "", false, true);
runTorchNet("net_lp_pooling_square", "", false, true);
runTorchNet("net_lp_pooling_power", "", false, true);
}
TEST_P(Test_Torch_layers, net_conv_gemm_lrn)
{
runTorchNet("net_conv_gemm_lrn", GetParam(), "", false, true);
if (backend == DNN_BACKEND_INFERENCE_ENGINE && target == DNN_TARGET_MYRIAD)
throw SkipTestException("");
runTorchNet("net_conv_gemm_lrn", "", false, true,
target == DNN_TARGET_OPENCL_FP16 ? 0.046 : 0.0,
target == DNN_TARGET_OPENCL_FP16 ? 0.023 : 0.0);
}
TEST_P(Test_Torch_layers, net_inception_block)
{
runTorchNet("net_inception_block", GetParam(), "", false, true);
runTorchNet("net_inception_block", "", false, true);
}
TEST_P(Test_Torch_layers, net_normalize)
{
runTorchNet("net_normalize", GetParam(), "", false, true);
runTorchNet("net_normalize", "", false, true);
}
TEST_P(Test_Torch_layers, net_padding)
{
int targetId = GetParam();
runTorchNet("net_padding", targetId, "", false, true);
runTorchNet("net_spatial_zero_padding", targetId, "", false, true);
runTorchNet("net_spatial_reflection_padding", targetId, "", false, true);
runTorchNet("net_padding", "", false, true);
runTorchNet("net_spatial_zero_padding", "", false, true);
runTorchNet("net_spatial_reflection_padding", "", false, true);
}
TEST_P(Test_Torch_layers, net_non_spatial)
{
runTorchNet("net_non_spatial", GetParam(), "", false, true);
if (backend == DNN_BACKEND_INFERENCE_ENGINE &&
(target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
throw SkipTestException("");
runTorchNet("net_non_spatial", "", false, true);
}
INSTANTIATE_TEST_CASE_P(/**/, Test_Torch_layers, availableDnnTargets());
TEST_P(Test_Torch_layers, run_paralel)
{
if (backend != DNN_BACKEND_OPENCV || target != DNN_TARGET_CPU)
throw SkipTestException("");
runTorchNet("net_parallel", "l5_torchMerge");
}
TEST_P(Test_Torch_layers, net_residual)
{
runTorchNet("net_residual", "", false, true);
}
typedef testing::TestWithParam<Target> Test_Torch_nets;
@@ -313,21 +347,6 @@ TEST_P(Test_Torch_nets, FastNeuralStyle_accuracy)
INSTANTIATE_TEST_CASE_P(/**/, Test_Torch_nets, availableDnnTargets());
// TODO: fix OpenCL and add to the rest of tests
TEST(Torch_Importer, run_paralel)
{
runTorchNet("net_parallel", DNN_TARGET_CPU, "l5_torchMerge");
}
TEST(Torch_Importer, DISABLED_run_paralel)
{
runTorchNet("net_parallel", DNN_TARGET_OPENCL, "l5_torchMerge");
}
TEST(Torch_Importer, net_residual)
{
runTorchNet("net_residual", DNN_TARGET_CPU, "", false, true);
}
// Test a custom layer
// https://github.com/torch/nn/blob/master/doc/convolution.md#nn.SpatialUpSamplingNearest
@@ -374,17 +393,29 @@ public:
}
}
virtual void forward(InputArrayOfArrays, OutputArrayOfArrays, OutputArrayOfArrays) CV_OVERRIDE {}
private:
int scale;
};
TEST(Torch_Importer, upsampling_nearest)
TEST_P(Test_Torch_layers, upsampling_nearest)
{
// Test a custom layer.
CV_DNN_REGISTER_LAYER_CLASS(SpatialUpSamplingNearest, SpatialUpSamplingNearestLayer);
runTorchNet("net_spatial_upsampling_nearest", DNN_TARGET_CPU, "", false, true);
try
{
runTorchNet("net_spatial_upsampling_nearest", "", false, true);
}
catch (...)
{
LayerFactory::unregisterLayer("SpatialUpSamplingNearest");
throw;
}
LayerFactory::unregisterLayer("SpatialUpSamplingNearest");
// Test an implemented layer.
runTorchNet("net_spatial_upsampling_nearest", "", false, true);
}
INSTANTIATE_TEST_CASE_P(/**/, Test_Torch_layers, dnnBackendsAndTargets());
}