potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 00:03:03 +04:00
Code Packages Releases Wiki Activity

Merge remote-tracking branch 'upstream/3.4' into merge-3.4

Browse Source
This commit is contained in:
Alexander Alekhin
2020-06-04 17:55:00 +00:00
parent c3e8a82c9c ef3844a177
commit 7722a2b8a8
35 changed files with 1061 additions and 236 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/src/darknet/darknet_io.cpp
+7
View File
@@ -658,6 +658,8 @@ namespace cv {
if (pad)
padding = kernel_size / 2;
// Cannot divide 0
CV_Assert(stride > 0);
CV_Assert(kernel_size > 0 && filters > 0);
CV_Assert(tensor_shape[0] > 0);
CV_Assert(tensor_shape[0] % groups == 0);
@@ -690,6 +692,9 @@ namespace cv {
int kernel_size = getParam<int>(layer_params, "size", 2);
int stride = getParam<int>(layer_params, "stride", 2);
int padding = getParam<int>(layer_params, "padding", kernel_size - 1);
// Cannot divide 0
CV_Assert(stride > 0);
setParams.setMaxpool(kernel_size, padding, stride);
tensor_shape[1] = (tensor_shape[1] - kernel_size + padding) / stride + 1;
@@ -732,6 +737,8 @@ namespace cv {
else if (layer_type == "reorg")
{
int stride = getParam<int>(layer_params, "stride", 2);
// Cannot divide 0
CV_Assert(stride > 0);
tensor_shape[0] = tensor_shape[0] * (stride * stride);
tensor_shape[1] = tensor_shape[1] / stride;
tensor_shape[2] = tensor_shape[2] / stride;
modules/dnn/src/dnn.cpp
+59 -22
View File
@@ -3508,6 +3508,7 @@ Net Net::Impl::createNetworkFromModelOptimizer(InferenceEngine::CNNNetwork& ieNe
for (auto& it : ieNet.getOutputsInfo())
{
CV_TRACE_REGION("output");
const auto& outputName = it.first;
LayerParams lp;
int lid = cvNet.addLayer(it.first, "", lp);
@@ -3517,37 +3518,60 @@ Net Net::Impl::createNetworkFromModelOptimizer(InferenceEngine::CNNNetwork& ieNe
#ifdef HAVE_DNN_NGRAPH
if (DNN_BACKEND_INFERENCE_ENGINE_NGRAPH == getInferenceEngineBackendTypeParam())
{
const auto& outputName = it.first;
Ptr<Layer> cvLayer(new NgraphBackendLayer(ieNet));
cvLayer->name = outputName;
cvLayer->type = "_unknown_";
if (ngraphFunction)
auto process_layer = [&](const std::string& name) -> bool
{
CV_TRACE_REGION("ngraph_function");
bool found = false;
for (const auto& op : ngraphOperations)
if (ngraphFunction)
{
CV_Assert(op);
if (op->get_friendly_name() == outputName)
CV_TRACE_REGION("ngraph_function");
for (const auto& op : ngraphOperations)
{
const std::string typeName = op->get_type_info().name;
cvLayer->type = typeName;
found = true;
break;
CV_Assert(op);
if (op->get_friendly_name() == name)
{
const std::string typeName = op->get_type_info().name;
cvLayer->type = typeName;
return true;
}
}
return false;
}
else
{
CV_TRACE_REGION("legacy_cnn_layer");
try
{
InferenceEngine::CNNLayerPtr ieLayer = ieNet.getLayerByName(name.c_str());
CV_Assert(ieLayer);
cvLayer->type = ieLayer->type;
return true;
}
catch (const std::exception& e)
{
CV_UNUSED(e);
CV_LOG_DEBUG(NULL, "IE layer extraction failure: '" << name << "' - " << e.what());
return false;
}
}
if (!found)
CV_LOG_WARNING(NULL, "DNN/IE: Can't determine output layer type: '" << outputName << "'");
}
else
{
CV_TRACE_REGION("legacy_cnn_layer");
InferenceEngine::CNNLayerPtr ieLayer = ieNet.getLayerByName(it.first.c_str());
CV_Assert(ieLayer);
};
cvLayer->type = ieLayer->type;
bool found = process_layer(outputName);
if (!found)
{
auto pos = outputName.rfind('.'); // cut port number: ".0"
if (pos != std::string::npos)
{
std::string layerName = outputName.substr(0, pos);
found = process_layer(layerName);
}
}
if (!found)
CV_LOG_WARNING(NULL, "DNN/IE: Can't determine output layer type: '" << outputName << "'");
ld.layerInstance = cvLayer;
ld.backendNodes[DNN_BACKEND_INFERENCE_ENGINE_NGRAPH] = backendNode;
}
@@ -3557,10 +3581,23 @@ Net Net::Impl::createNetworkFromModelOptimizer(InferenceEngine::CNNNetwork& ieNe
#ifdef HAVE_DNN_IE_NN_BUILDER_2019
Ptr<Layer> cvLayer(new InfEngineBackendLayer(ieNet));
InferenceEngine::CNNLayerPtr ieLayer = ieNet.getLayerByName(it.first.c_str());
InferenceEngine::CNNLayerPtr ieLayer;
try
{
ieLayer = ieNet.getLayerByName(outputName.c_str());
}
catch (...)
{
auto pos = outputName.rfind('.'); // cut port number: ".0"
if (pos != std::string::npos)
{
std::string layerName = outputName.substr(0, pos);
ieLayer = ieNet.getLayerByName(layerName.c_str());
}
}
CV_Assert(ieLayer);
cvLayer->name = it.first;
cvLayer->name = outputName;
cvLayer->type = ieLayer->type;
ld.layerInstance = cvLayer;
modules/dnn/src/onnx/onnx_importer.cpp
+4
View File
@@ -806,6 +806,10 @@ void ONNXImporter::populateNet(Net dstNet)
{
layerParams.type = "ELU";
}
else if (layer_type == "Tanh")
{
layerParams.type = "TanH";
}
else if (layer_type == "PRelu")
{
layerParams.type = "PReLU";
modules/dnn/test/test_tf_importer.cpp
+1 -1
View File
@@ -1220,7 +1220,7 @@ TEST_P(Test_TensorFlow_nets, EfficientDet)
}
checkBackend();
std::string proto = findDataFile("dnn/efficientdet-d0.pbtxt");
std::string model = findDataFile("dnn/efficientdet-d0.pb");
std::string model = findDataFile("dnn/efficientdet-d0.pb", false);
Net net = readNetFromTensorflow(model, proto);
Mat img = imread(findDataFile("dnn/dog416.png"));