potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 08:13:04 +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
2019-08-05 18:11:43 +00:00
parent 0757a51e2b 90ccd2f37c
commit 174b4ce29d
41 changed files with 423 additions and 302 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/src/caffe/opencv-caffe.proto
+2 -2
View File
@@ -116,9 +116,9 @@ message PriorBoxParameter {
CENTER_SIZE = 2;
}
// Minimum box size (in pixels). Required!
optional float min_size = 1;
repeated float min_size = 1;
// Maximum box size (in pixels). Required!
optional float max_size = 2;
repeated float max_size = 2;
// Various of aspect ratios. Duplicate ratios will be ignored.
// If none is provided, we use default ratio 1.
repeated float aspect_ratio = 3;
modules/dnn/src/layers/detection_output_layer.cpp
+2 -1
View File
@@ -198,7 +198,7 @@ public:
virtual bool supportBackend(int backendId) CV_OVERRIDE
{
return backendId == DNN_BACKEND_OPENCV ||
(backendId == DNN_BACKEND_INFERENCE_ENGINE && !_locPredTransposed && _bboxesNormalized && !_clip);
(backendId == DNN_BACKEND_INFERENCE_ENGINE && !_locPredTransposed && _bboxesNormalized);
}
bool getMemoryShapes(const std::vector<MatShape> &inputs,
@@ -936,6 +936,7 @@ public:
InferenceEngine::Builder::Layer l = ieLayer;
l.getParameters()["eta"] = std::string("1.0");
l.getParameters()["clip"] = _clip;
return Ptr<BackendNode>(new InfEngineBackendNode(l));
}
modules/dnn/src/layers/elementwise_layers.cpp
+1 -1
View File
@@ -796,7 +796,7 @@ struct AbsValFunctor
#ifdef HAVE_INF_ENGINE
InferenceEngine::Builder::Layer initInfEngineBuilderAPI()
{
return InferenceEngine::Builder::ReLULayer("").setNegativeSlope(-1);
return InferenceEngine::Builder::ReLULayer("").setNegativeSlope(-0.999999f);
}
#endif // HAVE_INF_ENGINE
modules/dnn/src/layers/prior_box_layer.cpp
+33 -31
View File
@@ -181,21 +181,20 @@ public:
PriorBoxLayerImpl(const LayerParams &params)
{
setParamsFrom(params);
_minSize = getParameter<float>(params, "min_size", 0, false, 0);
_flip = getParameter<bool>(params, "flip", 0, false, true);
_clip = getParameter<bool>(params, "clip", 0, false, true);
_bboxesNormalized = getParameter<bool>(params, "normalized_bbox", 0, false, true);
_aspectRatios.clear();
getParams("min_size", params, &_minSize);
getAspectRatios(params);
getVariance(params);
_maxSize = -1;
if (params.has("max_size"))
{
_maxSize = params.get("max_size").get<float>(0);
CV_Assert(_maxSize > _minSize);
getParams("max_size", params, &_maxSize);
CV_Assert(_minSize.size() == _maxSize.size());
for (int i = 0; i < _maxSize.size(); i++)
CV_Assert(_minSize[i] < _maxSize[i]);
}
std::vector<float> widths, heights;
@@ -214,25 +213,28 @@ public:
}
else
{
CV_Assert(_minSize > 0);
_boxWidths.resize(1 + (_maxSize > 0 ? 1 : 0) + _aspectRatios.size());
_boxHeights.resize(_boxWidths.size());
_boxWidths[0] = _boxHeights[0] = _minSize;
int i = 1;
if (_maxSize > 0)
CV_Assert(!_minSize.empty());
for (int i = 0; i < _minSize.size(); ++i)
{
// second prior: aspect_ratio = 1, size = sqrt(min_size * max_size)
_boxWidths[i] = _boxHeights[i] = sqrt(_minSize * _maxSize);
i += 1;
}
float minSize = _minSize[i];
CV_Assert(minSize > 0);
_boxWidths.push_back(minSize);
_boxHeights.push_back(minSize);
// rest of priors
for (size_t r = 0; r < _aspectRatios.size(); ++r)
{
float arSqrt = sqrt(_aspectRatios[r]);
_boxWidths[i + r] = _minSize * arSqrt;
_boxHeights[i + r] = _minSize / arSqrt;
if (_maxSize.size() > 0)
{
float size = sqrt(minSize * _maxSize[i]);
_boxWidths.push_back(size);
_boxHeights.push_back(size);
}
// rest of priors
for (size_t r = 0; r < _aspectRatios.size(); ++r)
{
float arSqrt = sqrt(_aspectRatios[r]);
_boxWidths.push_back(minSize * arSqrt);
_boxHeights.push_back(minSize / arSqrt);
}
}
}
CV_Assert(_boxWidths.size() == _boxHeights.size());
@@ -272,8 +274,9 @@ public:
virtual bool supportBackend(int backendId) CV_OVERRIDE
{
return backendId == DNN_BACKEND_OPENCV ||
(backendId == DNN_BACKEND_INFERENCE_ENGINE && haveInfEngine()) ||
(backendId == DNN_BACKEND_VKCOM && haveVulkan());
(backendId == DNN_BACKEND_INFERENCE_ENGINE && haveInfEngine() &&
( _explicitSizes || (_minSize.size() == 1 && _maxSize.size() <= 1)))
|| (backendId == DNN_BACKEND_VKCOM && haveVulkan());
}
bool getMemoryShapes(const std::vector<MatShape> &inputs,
@@ -523,10 +526,9 @@ public:
InferenceEngine::Builder::PriorBoxLayer ieLayer(name);
CV_Assert(!_explicitSizes);
ieLayer.setMinSize(_minSize);
if (_maxSize > 0)
ieLayer.setMaxSize(_maxSize);
ieLayer.setMinSize(_minSize[0]);
if (!_maxSize.empty())
ieLayer.setMaxSize(_maxSize[0]);
CV_CheckEQ(_offsetsX.size(), (size_t)1, ""); CV_CheckEQ(_offsetsY.size(), (size_t)1, ""); CV_CheckEQ(_offsetsX[0], _offsetsY[0], "");
ieLayer.setOffset(_offsetsX[0]);
@@ -573,8 +575,8 @@ public:
}
private:
float _minSize;
float _maxSize;
std::vector<float> _minSize;
std::vector<float> _maxSize;
float _stepX, _stepY;
modules/dnn/src/onnx/onnx_importer.cpp
+37 -4
View File
@@ -465,6 +465,20 @@ void ONNXImporter::populateNet(Net dstNet)
}
layerParams.set("begin", DictValue::arrayInt(&begin[0], begin.size()));
layerParams.set("end", DictValue::arrayInt(&end[0], end.size()));
}
else if (layer_type == "Split")
{
DictValue splits = layerParams.get("split");
const int numSplits = splits.size();
CV_Assert(numSplits > 1);
std::vector<int> slicePoints(numSplits - 1, splits.get<int>(0));
for (int i = 1; i < splits.size() - 1; ++i)
{
slicePoints[i] = slicePoints[i - 1] + splits.get<int>(i - 1);
}
layerParams.set("slice_point", DictValue::arrayInt(&slicePoints[0], slicePoints.size()));
layerParams.type = "Slice";
}
else if (layer_type == "Add" || layer_type == "Sum")
{
@@ -486,6 +500,11 @@ void ONNXImporter::populateNet(Net dstNet)
layerParams.type = "Eltwise";
}
}
else if (layer_type == "Max")
{
layerParams.type = "Eltwise";
layerParams.set("operation", "max");
}
else if (layer_type == "Sub")
{
Mat blob = getBlob(node_proto, constBlobs, 1);
@@ -741,6 +760,16 @@ void ONNXImporter::populateNet(Net dstNet)
{
layerParams.type = "Permute";
replaceLayerParam(layerParams, "perm", "order");
CV_Assert(node_proto.input_size() == 1);
if (constBlobs.find(node_proto.input(0)) != constBlobs.end())
{
std::vector<Mat> inputs(1, getBlob(node_proto, constBlobs, 0)), transposed;
runLayer(layerParams, inputs, transposed);
CV_Assert(transposed.size() == 1);
constBlobs.insert(std::make_pair(layerParams.name, transposed[0]));
continue;
}
}
else if (layer_type == "Unsqueeze")
{
@@ -906,8 +935,10 @@ void ONNXImporter::populateNet(Net dstNet)
}
int id = dstNet.addLayer(layerParams.name, layerParams.type, layerParams);
layer_id.insert(std::make_pair(layerParams.name, LayerInfo(id, 0)));
for (int i = 0; i < node_proto.output_size(); ++i)
{
layer_id.insert(std::make_pair(node_proto.output(i), LayerInfo(id, i)));
}
std::vector<MatShape> layerInpShapes, layerOutShapes, layerInternalShapes;
for (int j = 0; j < node_proto.input_size(); j++) {
@@ -924,8 +955,10 @@ void ONNXImporter::populateNet(Net dstNet)
// Compute shape of output blob for this layer.
Ptr<Layer> layer = dstNet.getLayer(id);
layer->getMemoryShapes(layerInpShapes, 0, layerOutShapes, layerInternalShapes);
CV_Assert(!layerOutShapes.empty());
outShapes[layerParams.name] = layerOutShapes[0];
for (int i = 0; i < node_proto.output_size() && i < (int)layerOutShapes.size(); ++i)
{
outShapes[node_proto.output(i)] = layerOutShapes[i];
}
}
}