mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2024-11-24 03:00:14 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #12264 from dkurt:dnn_remove_forward_method

Browse Source 
* Remove a forward method in dnn::Layer

* Add a test

* Fix tests

* Mark multiple dnn::Layer::finalize methods as deprecated

* Replace back dnn's inputBlobs to vector of pointers

* Remove Layer::forward_fallback from CV_OCL_RUN scopes
This commit is contained in:
Dmitry Kurtaev 2018-09-06 13:26:47 +03:00 committed by Alexander Alekhin
parent 8e8315abfd
commit d486204a0d
40 changed files with 820 additions and 487 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
modules/dnn/include/opencv2/dnn/dnn.hpp
View File

@ -46,9 +46,9 @@
#include <opencv2/core.hpp>
#if !defined CV_DOXYGEN && !defined CV_DNN_DONT_ADD_EXPERIMENTAL_NS
#define CV__DNN_EXPERIMENTAL_NS_BEGIN namespace experimental_dnn_34_v7 {
#define CV__DNN_EXPERIMENTAL_NS_BEGIN namespace experimental_dnn_34_v8 {
#define CV__DNN_EXPERIMENTAL_NS_END }
namespace cv { namespace dnn { namespace experimental_dnn_34_v7 { } using namespace experimental_dnn_34_v7; }}
namespace cv { namespace dnn { namespace experimental_dnn_34_v8 { } using namespace experimental_dnn_34_v8; }}
#else
#define CV__DNN_EXPERIMENTAL_NS_BEGIN
#define CV__DNN_EXPERIMENTAL_NS_END
@ -165,8 +165,6 @@ CV__DNN_EXPERIMENTAL_NS_BEGIN
};
class CV_EXPORTS ActivationLayer;
class CV_EXPORTS BatchNormLayer;
class CV_EXPORTS ScaleLayer;
/** @brief This interface class allows to build new Layers - are building blocks of networks.
*
@ -181,20 +179,31 @@ CV__DNN_EXPERIMENTAL_NS_BEGIN
CV_PROP_RW std::vector<Mat> blobs;
/** @brief Computes and sets internal parameters according to inputs, outputs and blobs.
* @deprecated Use Layer::finalize(InputArrayOfArrays, OutputArrayOfArrays) instead
* @param[in] input vector of already allocated input blobs
* @param[out] output vector of already allocated output blobs
*
* If this method is called after network has allocated all memory for input and output blobs
* and before inferencing.
*/
virtual void finalize(const std::vector<Mat*> &input, std::vector<Mat> &output);
CV_DEPRECATED virtual void finalize(const std::vector<Mat*> &input, std::vector<Mat> &output);
/** @brief Computes and sets internal parameters according to inputs, outputs and blobs.
* @param[in] inputs vector of already allocated input blobs
* @param[out] outputs vector of already allocated output blobs
*
* If this method is called after network has allocated all memory for input and output blobs
* and before inferencing.
*/
CV_WRAP virtual void finalize(InputArrayOfArrays inputs, OutputArrayOfArrays outputs);
/** @brief Given the @p input blobs, computes the output @p blobs.
* @deprecated Use Layer::forward(InputArrayOfArrays, OutputArrayOfArrays, OutputArrayOfArrays) instead
* @param[in] input the input blobs.
* @param[out] output allocated output blobs, which will store results of the computation.
* @param[out] internals allocated internal blobs
*/
virtual void forward(std::vector<Mat*> &input, std::vector<Mat> &output, std::vector<Mat> &internals) = 0;
CV_DEPRECATED virtual void forward(std::vector<Mat*> &input, std::vector<Mat> &output, std::vector<Mat> &internals);
/** @brief Given the @p input blobs, computes the output @p blobs.
* @param[in] inputs the input blobs.
@ -210,14 +219,22 @@ CV__DNN_EXPERIMENTAL_NS_BEGIN
*/
void forward_fallback(InputArrayOfArrays inputs, OutputArrayOfArrays outputs, OutputArrayOfArrays internals);
/** @brief @overload */
CV_WRAP void finalize(const std::vector<Mat> &inputs, CV_OUT std::vector<Mat> &outputs);
/** @brief
* @overload
* @deprecated Use Layer::finalize(InputArrayOfArrays, OutputArrayOfArrays) instead
*/
CV_DEPRECATED void finalize(const std::vector<Mat> &inputs, CV_OUT std::vector<Mat> &outputs);
/** @brief @overload */
CV_WRAP std::vector<Mat> finalize(const std::vector<Mat> &inputs);
/** @brief
* @overload
* @deprecated Use Layer::finalize(InputArrayOfArrays, OutputArrayOfArrays) instead
*/
CV_DEPRECATED std::vector<Mat> finalize(const std::vector<Mat> &inputs);
/** @brief Allocates layer and computes output. */
CV_WRAP void run(const std::vector<Mat> &inputs, CV_OUT std::vector<Mat> &outputs,
/** @brief Allocates layer and computes output.
* @deprecated This method will be removed in the future release.
*/
CV_DEPRECATED CV_WRAP void run(const std::vector<Mat> &inputs, CV_OUT std::vector<Mat> &outputs,
CV_IN_OUT std::vector<Mat> &internals);
/** @brief Returns index of input blob into the input array.
@ -388,9 +405,6 @@ CV__DNN_EXPERIMENTAL_NS_BEGIN
/** @brief Returns pointers to input layers of specific layer. */
std::vector<Ptr<Layer> > getLayerInputs(LayerId layerId); // FIXIT: CV_WRAP
/** @brief Delete layer for the network (not implemented yet) */
CV_WRAP void deleteLayer(LayerId layer);
/** @brief Connects output of the first layer to input of the second layer.
* @param outPin descriptor of the first layer output.
* @param inpPin descriptor of the second layer input.

15
modules/dnn/misc/python/pyopencv_dnn.hpp
View File

@ -146,16 +146,16 @@ public:
return false;
}
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &) CV_OVERRIDE
virtual void forward(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
PyGILState_STATE gstate;
gstate = PyGILState_Ensure();
std::vector<Mat> inps(inputs.size());
for (size_t i = 0; i < inputs.size(); ++i)
inps[i] = *inputs[i];
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
PyObject* args = pyopencv_from(inps);
PyObject* args = pyopencv_from(inputs);
PyObject* res = PyObject_CallMethodObjArgs(o, PyString_FromString("forward"), args, NULL);
Py_DECREF(args);
PyGILState_Release(gstate);
@ -174,11 +174,6 @@ public:
}
}
virtual void forward(InputArrayOfArrays, OutputArrayOfArrays, OutputArrayOfArrays) CV_OVERRIDE
{
CV_Error(Error::StsNotImplemented, "");
}
private:
// Map layers types to python classes.
static std::map<std::string, std::vector<PyObject*> > pyLayers;

73
modules/dnn/src/dnn.cpp
View File

@ -430,19 +430,24 @@ struct DataLayer : public Layer
backendId == DNN_BACKEND_INFERENCE_ENGINE && inputsData.size() == 1;
}
void forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs, OutputArrayOfArrays internals) CV_OVERRIDE
void forward(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays internals_arr) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget),
forward_ocl(inputs, outputs, internals));
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs, outputs, internals);
if (outputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*>&, std::vector<Mat>& outputs, std::vector<Mat> &) CV_OVERRIDE
{
std::vector<Mat> outputs, internals;
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
// Supported modes:
// | Input type | Output type |
// | fp32 | fp32 |
@ -567,8 +572,11 @@ struct DataLayer : public Layer
return false;
}
void finalize(const std::vector<Mat*>&, std::vector<Mat>& outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<Mat> outputs;
outputs_arr.getMatVector(outputs);
CV_Assert_N(outputs.size() == scaleFactors.size(), outputs.size() == means.size(),
inputsData.size() == outputs.size());
skip = true;
@ -1414,6 +1422,7 @@ struct Net::Impl
addInfEngineNetOutputs(ld);
net = Ptr<InfEngineBackendNet>();
netBlobsWrappers.clear();
layer->preferableTarget = DNN_TARGET_CPU;
continue;
}
ld.skip = true; // Initially skip all Inference Engine supported layers.
@ -1622,7 +1631,12 @@ struct Net::Impl
Ptr<Layer> layerPtr = ld.getLayerInstance();
{
layerPtr->finalize(ld.inputBlobs, ld.outputBlobs);
std::vector<Mat> inps(ld.inputBlobs.size());
for (int i = 0; i < ld.inputBlobs.size(); ++i)
{
inps[i] = *ld.inputBlobs[i];
}
layerPtr->finalize(inps, ld.outputBlobs);
layerPtr->preferableTarget = preferableTarget;
#if 0
std::cout << "\toutputs:";
@ -2138,7 +2152,12 @@ struct Net::Impl
ld.inputBlobsWrappers[i]->copyToHost();
}
layer->forward(ld.inputBlobs, ld.outputBlobs, ld.internals);
std::vector<Mat> inps(ld.inputBlobs.size());
for (int i = 0; i < ld.inputBlobs.size(); ++i)
{
inps[i] = *ld.inputBlobs[i];
}
layer->forward(inps, ld.outputBlobs, ld.internals);
if (DNN_CHECK_NAN_INF)
{
@ -2712,11 +2731,6 @@ int Net::getLayerId(const String &layer)
return impl->getLayerId(layer);
}
void Net::deleteLayer(LayerId)
{
CV_Error(Error::StsNotImplemented, "");
}
Ptr<Layer> Net::getLayer(LayerId layerId)
{
LayerData &ld = impl->getLayerData(layerId);
@ -3172,10 +3186,7 @@ static void vecToPVec(const std::vector<T> &v, std::vector<T*> &pv)
void Layer::finalize(const std::vector<Mat> &inputs, std::vector<Mat> &outputs)
{
CV_TRACE_FUNCTION();
std::vector<Mat*> inputsp;
vecToPVec(inputs, inputsp);
this->finalize(inputsp, outputs);
this->finalize((InputArrayOfArrays)inputs, (OutputArrayOfArrays)outputs);
}
void Layer::finalize(const std::vector<Mat*> &input, std::vector<Mat> &output)
@ -3183,6 +3194,18 @@ void Layer::finalize(const std::vector<Mat*> &input, std::vector<Mat> &output)
(void)input;(void)output;
}
void Layer::finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr)
{
CV_TRACE_FUNCTION();
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
std::vector<Mat*> inputsp;
vecToPVec(inputs, inputsp);
this->finalize(inputsp, outputs);
}
std::vector<Mat> Layer::finalize(const std::vector<Mat> &inputs)
{
CV_TRACE_FUNCTION();
@ -3192,12 +3215,17 @@ std::vector<Mat> Layer::finalize(const std::vector<Mat> &inputs)
return outputs;
}
void Layer::forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs, OutputArrayOfArrays internals)
void Layer::forward(std::vector<Mat*> &input, std::vector<Mat> &output, std::vector<Mat> &internals)
{
// We kept this method for compatibility. DNN calls it now only to support users' implementations.
}
void Layer::forward(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays internals_arr)
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs, outputs, internals);
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
}
void Layer::forward_fallback(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays internals_arr)
@ -3241,7 +3269,6 @@ void Layer::forward_fallback(InputArrayOfArrays inputs_arr, OutputArrayOfArrays
internals_arr.assign(orig_internals);
return;
}
std::vector<Mat> inpvec;
std::vector<Mat> outputs;
std::vector<Mat> internals;
@ -3265,10 +3292,8 @@ void Layer::run(const std::vector<Mat> &inputs, std::vector<Mat> &outputs, std::
{
CV_TRACE_FUNCTION();
std::vector<Mat*> inputsp;
vecToPVec(inputs, inputsp);
this->finalize(inputsp, outputs);
this->forward(inputsp, outputs, internals);
this->finalize(inputs, outputs);
this->forward(inputs, outputs, internals);
}
Layer::~Layer() {}

14
modules/dnn/src/layers/batch_norm_layer.cpp
View File

@ -234,18 +234,20 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(blobs.size() >= 2);
CV_Assert(inputs.size() == 1);
Mat &inpBlob = *inputs[0];
Mat &inpBlob = inputs[0];
CV_Assert(inpBlob.dims == 2 || inpBlob.dims == 4);
int rows = inpBlob.dims > 2 ? inpBlob.size[2] : 1;
int cols = inpBlob.dims > 2 ? inpBlob.size[3] : 1;

16
modules/dnn/src/layers/blank_layer.cpp
View File

@ -99,17 +99,19 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
for (int i = 0, n = outputs.size(); i < n; ++i)
if (outputs[i].data != inputs[i]->data)
inputs[i]->copyTo(outputs[i]);
if (outputs[i].data != inputs[i].data)
inputs[i].copyTo(outputs[i]);
}
virtual Ptr<BackendNode> initInfEngine(const std::vector<Ptr<BackendWrapper> >&) CV_OVERRIDE

30
modules/dnn/src/layers/concat_layer.cpp
View File

@ -111,12 +111,12 @@ public:
class ChannelConcatInvoker : public ParallelLoopBody
{
public:
std::vector<Mat*>* inputs;
std::vector<Mat>* inputs;
Mat* output;
int nstripes;
std::vector<const float*> chptrs;
static void run(std::vector<Mat*>& inputs, Mat& output, int nstripes)
static void run(std::vector<Mat>& inputs, Mat& output, int nstripes)
{
ChannelConcatInvoker cc;
cc.inputs = &inputs;
@ -127,7 +127,7 @@ public:
int nchannels = 0, batchsz = output.size[0];
for( i = 0; i < ninputs; i++ )
{
Mat& inp = *inputs[i];
Mat& inp = inputs[i];
CV_Assert( inp.isContinuous() && (inp.type() == CV_32F || inp.type() == CV_16S) &&
inp.dims == 4 && inp.size[0] == output.size[0] &&
inp.size[2] == output.size[2] &&
@ -142,7 +142,7 @@ public:
int ofs = 0;
for( i = 0; i < ninputs; i++)
{
Mat& inp = *inputs[i];
Mat& inp = inputs[i];
for( int j = 0; j < batchsz; j++ )
for( int k = 0; k < inp.size[1]; k++ )
{
@ -241,15 +241,17 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
int cAxis = clamp(axis, inputs[0]->dims);
int cAxis = clamp(axis, inputs[0].dims);
Mat& outMat = outputs[0];
if (padding)
@ -267,14 +269,14 @@ public:
ranges[cAxis].start = 0;
for (size_t i = 0; i < inputs.size(); i++)
{
ranges[cAxis].end = ranges[cAxis].start + inputs[i]->size[cAxis];
ranges[cAxis].end = ranges[cAxis].start + inputs[i].size[cAxis];
for (int j = 0; j < outMat.dims; ++j)
{
if (j == cAxis) continue;
ranges[j].start = (outMat.size[j] - inputs[i]->size[j]) / 2;
ranges[j].end = ranges[j].start + inputs[i]->size[j];
ranges[j].start = (outMat.size[j] - inputs[i].size[j]) / 2;
ranges[j].end = ranges[j].start + inputs[i].size[j];
}
inputs[i]->copyTo(outMat(&ranges[0]));
inputs[i].copyTo(outMat(&ranges[0]));
ranges[cAxis].start = ranges[cAxis].end;
}
}

129
modules/dnn/src/layers/convolution_layer.cpp
View File

@ -79,49 +79,24 @@ public:
adjustPad.height < stride.height);
}
virtual bool supportBackend(int backendId) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
#ifdef HAVE_INF_ENGINE
if (backendId == DNN_BACKEND_INFERENCE_ENGINE)
{
if (type == "Convolution")
return preferableTarget != DNN_TARGET_MYRIAD || dilation.width == dilation.height;
else
{
CV_Assert(type == "Deconvolution");
const int outGroupCn = blobs[0].size[1]; // Weights are in IOHW layout
const int group = numOutput / outGroupCn;
if (group != 1)
{
#if INF_ENGINE_VER_MAJOR_GE(INF_ENGINE_RELEASE_2018R3)
return preferableTarget == DNN_TARGET_CPU;
#endif
return false;
}
if (preferableTarget == DNN_TARGET_OPENCL || preferableTarget == DNN_TARGET_OPENCL_FP16)
return dilation.width == 1 && dilation.height == 1;
return true;
}
}
else
#endif // HAVE_INF_ENGINE
return backendId == DNN_BACKEND_OPENCV || backendId == DNN_BACKEND_HALIDE;
}
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
{
CV_Assert(inputs.size() > 0);
CV_Assert(blobs.size() >= 1 && blobs.size() <= 2);
CV_Assert(blobs[0].dims == 4 && blobs[0].size[3] == kernel.width && blobs[0].size[2] == kernel.height);
const Mat &input = *inputs[0];
const Mat &input = inputs[0];
CV_Assert(input.dims == 4 && (input.type() == CV_32F || input.type() == CV_64F || input.type() == CV_16S));
for (size_t i = 0; i < inputs.size(); i++)
{
CV_Assert(inputs[i]->type() == input.type());
CV_Assert(inputs[i]->dims == 4 && inputs[i]->size[1] == input.size[1]);
CV_Assert(inputs[i]->size[2] == input.size[2] && inputs[i]->size[3] == input.size[3]);
CV_Assert(inputs[i].type() == input.type());
CV_Assert(inputs[i].dims == 4 && inputs[i].size[1] == input.size[1]);
CV_Assert(inputs[i].size[2] == input.size[2] && inputs[i].size[3] == input.size[3]);
}
Size outSize = Size(outputs[0].size[3], outputs[0].size[2]);
@ -225,6 +200,14 @@ public:
return shape(out.area(), ksize);
}
virtual bool supportBackend(int backendId) CV_OVERRIDE
{
if (backendId == DNN_BACKEND_INFERENCE_ENGINE)
return preferableTarget != DNN_TARGET_MYRIAD || dilation.width == dilation.height;
else
return backendId == DNN_BACKEND_OPENCV || backendId == DNN_BACKEND_HALIDE;
}
bool getMemoryShapes(const std::vector<MatShape> &inputs,
const int requiredOutputs,
std::vector<MatShape> &outputs,
@ -262,9 +245,9 @@ public:
return false;
}
virtual void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
BaseConvolutionLayerImpl::finalize(inputs, outputs);
BaseConvolutionLayerImpl::finalize(inputs_arr, outputs_arr);
CV_Assert(!blobs.empty());
const int outCn = blobs[0].size[0];
@ -1007,22 +990,24 @@ public:
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
/*printf("conv %s: input (%d x %d x %d x %d), kernel (%d x %d), pad (%d x %d), stride (%d x %d), dilation (%d x %d)\n",
name.c_str(), inputs[0]->size[0], inputs[0]->size[1], inputs[0]->size[2], inputs[0]->size[3],
name.c_str(), inputs[0].size[0], inputs[0].size[1], inputs[0].size[2], inputs[0].size[3],
kernel.width, kernel.height, pad.width, pad.height,
stride.width, stride.height, dilation.width, dilation.height);*/
CV_Assert_N(inputs.size() == (size_t)1, inputs[0]->size[1] % blobs[0].size[1] == 0,
outputs.size() == 1, inputs[0]->data != outputs[0].data);
CV_Assert_N(inputs.size() == (size_t)1, inputs[0].size[1] % blobs[0].size[1] == 0,
outputs.size() == 1, inputs[0].data != outputs[0].data);
int ngroups = inputs[0]->size[1]/blobs[0].size[1];
int ngroups = inputs[0].size[1]/blobs[0].size[1];
CV_Assert(outputs[0].size[1] % ngroups == 0);
int outCn = blobs[0].size[0];
@ -1049,7 +1034,7 @@ public:
int nstripes = std::max(getNumThreads(), 1);
ParallelConv::run(*inputs[0], outputs[0], weightsMat, biasvec, reluslope,
ParallelConv::run(inputs[0], outputs[0], weightsMat, biasvec, reluslope,
kernel, pad, stride, dilation, activ.get(), ngroups, nstripes);
}
@ -1089,6 +1074,29 @@ public:
return shape(ksize, inpH * inpW);
}
virtual bool supportBackend(int backendId) CV_OVERRIDE
{
#ifdef HAVE_INF_ENGINE
if (backendId == DNN_BACKEND_INFERENCE_ENGINE)
{
const int outGroupCn = blobs[0].size[1]; // Weights are in IOHW layout
const int group = numOutput / outGroupCn;
if (group != 1)
{
#if INF_ENGINE_VER_MAJOR_GE(INF_ENGINE_RELEASE_2018R3)
return preferableTarget == DNN_TARGET_CPU;
#endif
return false;
}
if (preferableTarget == DNN_TARGET_OPENCL || preferableTarget == DNN_TARGET_OPENCL_FP16)
return dilation.width == 1 && dilation.height == 1;
return true;
}
else
#endif // HAVE_INF_ENGINE
return backendId == DNN_BACKEND_OPENCV || backendId == DNN_BACKEND_HALIDE;
}
bool getMemoryShapes(const std::vector<MatShape> &inputs,
const int requiredOutputs,
std::vector<MatShape> &outputs,
@ -1141,11 +1149,15 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
BaseConvolutionLayerImpl::finalize(inputs, outputs);
BaseConvolutionLayerImpl::finalize(inputs_arr, outputs_arr);
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
getConvPoolPaddings(Size(outputs[0].size[3], outputs[0].size[2]),
Size(inputs[0]->size[3], inputs[0]->size[2]),
Size(inputs[0].size[3], inputs[0].size[2]),
kernel, stride, padMode, dilation, pad);
}
@ -1494,18 +1506,21 @@ public:
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget) &&
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
forward_ocl(inputs_arr, outputs_arr, internals_arr));
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat *> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
int outCn = numOutput;
int inpCn = inputs[0]->size[1];
int inpCn = inputs[0].size[1];
bool is1x1flag = is1x1();
int nstripes = getNumThreads();
@ -1520,13 +1535,13 @@ public:
int ngroups = outCn / blobs[0].size[1];
int inpGroupCn = inpCn / ngroups;
int outGroupCn = blobs[0].size[1];
const Mat& inp = *inputs[ii];
const Mat& inp = inputs[ii];
Mat& out = outputs[ii];
int numImg = inp.size[0];
int inpH = inp.size[2], inpW = inp.size[3];
int outH = out.size[2], outW = out.size[3];
Mat convBlob = inputs[ii]->reshape(1, numImg*inpCn);
Mat convBlob = inputs[ii].reshape(1, numImg*inpCn);
Mat decnBlob = out.reshape(1, numImg*outCn);
for (int n = 0; n < numImg; n++)

16
modules/dnn/src/layers/crop_and_resize_layer.cpp
View File

@ -40,17 +40,19 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
Mat& inp = *inputs[0];
Mat& inp = inputs[0];
Mat& out = outputs[0];
Mat boxes = inputs[1]->reshape(1, inputs[1]->total() / 7);
Mat boxes = inputs[1].reshape(1, inputs[1].total() / 7);
const int numChannels = inp.size[1];
const int inpHeight = inp.size[2];
const int inpWidth = inp.size[3];

26
modules/dnn/src/layers/crop_layer.cpp
View File

@ -90,12 +90,14 @@ public:
return false;
}
void finalize(const std::vector<Mat *> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
CV_Assert(2 == inputs.size());
const Mat &inpBlob = *inputs[0];
const Mat &inpSzBlob = *inputs[1];
const Mat &inpBlob = inputs[0];
const Mat &inpSzBlob = inputs[1];
int dims = inpBlob.dims;
int start_axis = clamp(startAxis, dims);
@ -135,18 +137,18 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat *> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
Mat &input = *inputs[0];
Mat &output = outputs[0];
input(&crop_ranges[0]).copyTo(output);
Mat &input = inputs[0];
input(&crop_ranges[0]).copyTo(outputs[0]);
}
virtual Ptr<BackendNode> initInfEngine(const std::vector<Ptr<BackendWrapper> >&) CV_OVERRIDE

31
modules/dnn/src/layers/detection_output_layer.cpp
View File

@ -419,27 +419,28 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
}
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
std::vector<LabelBBox> allDecodedBBoxes;
std::vector<Mat> allConfidenceScores;
int num = inputs[0]->size[0];
int num = inputs[0].size[0];
// extract predictions from input layers
{
int numPriors = inputs[2]->size[2] / 4;
int numPriors = inputs[2].size[2] / 4;
const float* locationData = inputs[0]->ptr<float>();
const float* confidenceData = inputs[1]->ptr<float>();
const float* priorData = inputs[2]->ptr<float>();
const float* locationData = inputs[0].ptr<float>();
const float* confidenceData = inputs[1].ptr<float>();
const float* priorData = inputs[2].ptr<float>();
// Retrieve all location predictions
std::vector<LabelBBox> allLocationPredictions;
@ -465,9 +466,9 @@ public:
else
{
// Input image sizes;
CV_Assert(inputs[3]->dims == 4);
clipBounds.xmax = inputs[3]->size[3] - 1;
clipBounds.ymax = inputs[3]->size[2] - 1;
CV_Assert(inputs[3].dims == 4);
clipBounds.xmax = inputs[3].size[3] - 1;
clipBounds.ymax = inputs[3].size[2] - 1;
}
}
DecodeBBoxesAll(allLocationPredictions, priorBBoxes, priorVariances, num,
@ -502,6 +503,8 @@ public:
allIndices[i], _groupByClasses);
}
CV_Assert(count == numKept);
// Sync results back due changed output shape.
outputs_arr.assign(outputs);
}
size_t outputDetections_(

11
modules/dnn/src/layers/elementwise_layers.cpp
View File

@ -187,16 +187,19 @@ public:
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(this->preferableTarget),
func.applyOCL(inputs_arr, outputs_arr, internals_arr))
if (inputs_arr.depth() == CV_16S)
{
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
for (size_t i = 0; i < inputs.size(); i++)
{
const Mat &src = *inputs[i];
const Mat &src = inputs[i];
Mat &dst = outputs[i];
CV_Assert(src.size == dst.size && src.type() == dst.type() &&
src.isContinuous() && dst.isContinuous() && src.type() == CV_32F);

34
modules/dnn/src/layers/eltwise_layer.cpp
View File

@ -123,7 +123,7 @@ public:
class EltwiseInvoker : public ParallelLoopBody
{
public:
const Mat** srcs;
const Mat* srcs;
int nsrcs;
Mat* dst;
const std::vector<float>* coeffs;
@ -135,7 +135,7 @@ public:
EltwiseInvoker() : srcs(0), nsrcs(0), dst(0), coeffs(0), op(PROD), nstripes(0), activ(0), channels(0), planeSize(0) {}
static void run(const Mat** srcs, int nsrcs, Mat& dst,
static void run(const Mat* srcs, int nsrcs, Mat& dst,
const std::vector<float>& coeffs, EltwiseOp op,
const ActivationLayer* activ, int nstripes)
{
@ -144,9 +144,9 @@ public:
for( int i = 0; i > nsrcs; i++ )
{
CV_Assert(srcs[i]->size == dst.size &&
srcs[i]->type() == dst.type() &&
srcs[i]->isContinuous());
CV_Assert(srcs[i].size == dst.size &&
srcs[i].type() == dst.type() &&
srcs[i].isContinuous());
}
EltwiseInvoker p;
@ -200,14 +200,14 @@ public:
for( c = 0; c < channels; c++ )
{
size_t globalDelta = delta + (sampleIdx*channels + c)*planeSize;
const float* srcptr0 = srcs[0]->ptr<float>() + globalDelta;
const float* srcptr0 = srcs[0].ptr<float>() + globalDelta;
float* dstptr = dstptr0 + globalDelta;
if( op == PROD )
{
for( k = 1; k < n; k++ )
{
const float* srcptr1 = srcs[k]->ptr<float>() + globalDelta;
const float* srcptr1 = srcs[k].ptr<float>() + globalDelta;
for( j = 0; j < blockSize; j++ )
{
dstptr[j] = srcptr0[j]*srcptr1[j];
@ -219,7 +219,7 @@ public:
{
for( k = 1; k < n; k++ )
{
const float* srcptr1 = srcs[k]->ptr<float>() + globalDelta;
const float* srcptr1 = srcs[k].ptr<float>() + globalDelta;
for( j = 0; j < blockSize; j++ )
{
dstptr[j] = std::max(srcptr0[j], srcptr1[j]);
@ -231,7 +231,7 @@ public:
{
for( k = 1; k < n; k++ )
{
const float* srcptr1 = srcs[k]->ptr<float>() + globalDelta;
const float* srcptr1 = srcs[k].ptr<float>() + globalDelta;
for( j = 0; j < blockSize; j++ )
{
dstptr[j] = srcptr0[j] + srcptr1[j];
@ -244,7 +244,7 @@ public:
float c0 = coeffsptr[0];
for( k = 1; k < n; k++ )
{
const float* srcptr1 = srcs[k]->ptr<float>() + globalDelta;
const float* srcptr1 = srcs[k].ptr<float>() + globalDelta;
float c1 = coeffsptr[k];
for( j = 0; j < blockSize; j++ )
{
@ -358,17 +358,19 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat *> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(outputs.size() == 1);
const int nstripes = getNumThreads();
EltwiseInvoker::run((const Mat**)&inputs[0], (int)inputs.size(), outputs[0],
EltwiseInvoker::run(&inputs[0], (int)inputs.size(), outputs[0],
coeffs, op, activ.get(), nstripes);
}

17
modules/dnn/src/layers/flatten_layer.cpp
View File

@ -139,18 +139,23 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
for (size_t i = 0; i < inputs.size(); i++)
{
MatShape outShape = shape(outputs[i]);
outputs[i] = inputs[i]->reshape(1, (int)outShape.size(), &outShape[0]);
if (inputs[i].data != outputs[i].data)
{
inputs[i].reshape(1, (int)outShape.size(), &outShape[0]).copyTo(outputs[i]);
}
}
}

20
modules/dnn/src/layers/fully_connected_layer.cpp
View File

@ -273,7 +273,7 @@ public:
};
#ifdef HAVE_OPENCL
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays, OutputArrayOfArrays) CV_OVERRIDE
{
innerProductOp.release();
}
@ -393,20 +393,22 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &input, std::vector<Mat> &output, std::vector<Mat> &) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> input, output;
inputs_arr.getMatVector(input);
outputs_arr.getMatVector(output);
int axisCan = clamp(axis, input[0]->dims);
int outerSize = input[0]->total(0, axisCan);
int axisCan = clamp(axis, input[0].dims);
int outerSize = input[0].total(0, axisCan);
for (size_t i = 0; i < input.size(); i++)
{
Mat srcMat = input[i]->reshape(1, outerSize);
Mat srcMat = input[i].reshape(1, outerSize);
Mat dstMat = output[i].reshape(1, outerSize);
const int nstripes = getNumThreads();

18
modules/dnn/src/layers/lrn_layer.cpp
View File

@ -96,7 +96,7 @@ public:
}
#ifdef HAVE_OPENCL
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays, OutputArrayOfArrays) CV_OVERRIDE
{
lrnOp.release();
}
@ -152,21 +152,23 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(inputs.size() == outputs.size());
for (int i = 0; i < inputs.size(); i++)
{
CV_Assert(inputs[i]->dims == 4);
CV_Assert(inputs[i].dims == 4);
Mat &src = *inputs[i];
Mat &src = inputs[i];
Mat &dst = outputs[i];
switch (type)

16
modules/dnn/src/layers/max_unpooling_layer.cpp
View File

@ -62,17 +62,19 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(inputs.size() == 2);
Mat& input = *inputs[0];
Mat& indices = *inputs[1];
Mat& input = inputs[0];
Mat& indices = inputs[1];
CV_Assert(input.total() == indices.total());
CV_Assert(input.size[0] == 1);

23
modules/dnn/src/layers/mvn_layer.cpp
View File

@ -96,13 +96,15 @@ public:
return fuse_relu;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
int splitDim = (acrossChannels) ? 1 : 2;
int i, newRows = 1;
for( i = 0; i < splitDim; i++ )
newRows *= inputs[0]->size[i];
zeroDev = inputs[0]->total() == newRows;
newRows *= inputs[0].size[i];
zeroDev = inputs[0].total() == newRows;
}
virtual bool supportBackend(int backendId) CV_OVERRIDE
@ -271,17 +273,20 @@ public:
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat *> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
for (size_t inpIdx = 0; inpIdx < inputs.size(); inpIdx++)
{
Mat &inpBlob = *inputs[inpIdx];
Mat &inpBlob = inputs[inpIdx];
Mat &outBlob = outputs[inpIdx];
int splitDim = (acrossChannels) ? 1 : 2;

27
modules/dnn/src/layers/normalize_bbox_layer.cpp
View File

@ -89,12 +89,14 @@ public:
return true;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
CV_Assert(inputs.size() == 1);
endAxis = endAxis == -1 ? (inputs[0]->dims - 1) : endAxis;
startAxis = startAxis == -1 ? (inputs[0]->dims - 1) : startAxis;
acrossSpatial = (startAxis == 1 && endAxis == inputs[0]->dims - 1);
endAxis = endAxis == -1 ? (inputs[0].dims - 1) : endAxis;
startAxis = startAxis == -1 ? (inputs[0].dims - 1) : startAxis;
acrossSpatial = (startAxis == 1 && endAxis == inputs[0].dims - 1);
}
#ifdef HAVE_OPENCL
@ -186,18 +188,21 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
CV_Assert(inputs.size() == 1 && outputs.size() == 1);
CV_Assert(inputs[0]->total() == outputs[0].total());
CV_Assert(inputs[0].total() == outputs[0].total());
const Mat& inp0 = *inputs[0];
const Mat& inp0 = inputs[0];
Mat& buffer = internals[0];
startAxis = clamp(startAxis, inp0.dims);
endAxis = clamp(endAxis, inp0.dims);

39
modules/dnn/src/layers/padding_layer.cpp
View File

@ -61,14 +61,17 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
// Compute dstRanges.
const MatSize& inpShape = inputs[0]->size;
const MatSize& inpShape = inputs[0].size;
dstRanges.resize(paddings.size());
int offset = 0;
if (inputDims != -1 && inputs[0]->dims != inputDims)
if (inputDims != -1 && inputs[0].dims != inputDims)
{
dstRanges.insert(dstRanges.begin(), Range::all());
offset = 1;
@ -81,7 +84,7 @@ public:
}
// Add the rest of dimensions.
for (int i = dstRanges.size(); i < inputs[0]->dims; ++i)
for (int i = dstRanges.size(); i < inputs[0].dims; ++i)
dstRanges.push_back(Range::all());
}
@ -96,31 +99,33 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
if (paddingType == "constant")
{
outputs[0].setTo(paddingValue);
inputs[0]->copyTo(outputs[0](dstRanges));
inputs[0].copyTo(outputs[0](dstRanges));
}
else if (paddingType == "reflect")
{
CV_Assert(inputs.size() == 1);
CV_Assert(outputs.size() == 1);
CV_Assert(inputs[0]->dims == 4);
CV_Assert(inputs[0].dims == 4);
CV_Assert(outputs[0].dims == 4);
if (inputs[0]->size[0] != outputs[0].size[0] || inputs[0]->size[1] != outputs[0].size[1])
if (inputs[0].size[0] != outputs[0].size[0] || inputs[0].size[1] != outputs[0].size[1])
CV_Error(Error::StsNotImplemented, "Only spatial reflection padding is supported.");
const int inpHeight = inputs[0]->size[2];
const int inpWidth = inputs[0]->size[3];
const int inpHeight = inputs[0].size[2];
const int inpWidth = inputs[0].size[3];
const int outHeight = outputs[0].size[2];
const int outWidth = outputs[0].size[3];
const int padTop = dstRanges[2].start;
@ -130,11 +135,11 @@ public:
CV_CheckLT(padTop, inpHeight, ""); CV_CheckLT(padBottom, inpHeight, "");
CV_CheckLT(padLeft, inpWidth, ""); CV_CheckLT(padRight, inpWidth, "");
for (size_t n = 0; n < inputs[0]->size[0]; ++n)
for (size_t n = 0; n < inputs[0].size[0]; ++n)
{
for (size_t ch = 0; ch < inputs[0]->size[1]; ++ch)
for (size_t ch = 0; ch < inputs[0].size[1]; ++ch)
{
copyMakeBorder(getPlane(*inputs[0], n, ch),
copyMakeBorder(getPlane(inputs[0], n, ch),
getPlane(outputs[0], n, ch),
padTop, padBottom, padLeft, padRight,
BORDER_REFLECT_101);

31
modules/dnn/src/layers/permute_layer.cpp
View File

@ -172,18 +172,21 @@ public:
_count = _oldStride[0] * shapeBefore[0];
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
if(!_needsPermute)
{
return;
}
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(inputs.size() > 0);
const Mat& inp0 = *inputs[0];
const Mat& inp0 = inputs[0];
CV_Assert((int)_numAxes == inp0.dims);
computeStrides(shape(*inputs[0]), shape(outputs[0]));
computeStrides(shape(inputs[0]), shape(outputs[0]));
#ifdef HAVE_OPENCL
if (uorder.empty())
@ -319,22 +322,24 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
size_t k, ninputs = inputs.size();
if(!_needsPermute)
{
for (k = 0; k < ninputs; k++)
{
CV_Assert(outputs[k].total() == inputs[k]->total());
if (outputs[k].data != inputs[k]->data)
inputs[k]->copyTo(outputs[k]);
CV_Assert(outputs[k].total() == inputs[k].total());
if (outputs[k].data != inputs[k].data)
inputs[k].copyTo(outputs[k]);
}
}
else
@ -346,10 +351,10 @@ public:
for (k = 0; k < ninputs; k++)
{
const Mat& inp = *inputs[k];
const Mat& inp = inputs[k];
Mat& out = outputs[k];
CV_Assert(inp.dims == numAxes && inp.size == inputs[0]->size);
CV_Assert(inp.dims == numAxes && inp.size == inputs[0].size);
CV_Assert(out.dims == numAxes && out.size == outputs[0].size);
CV_Assert(inp.isContinuous() && out.isContinuous());

27
modules/dnn/src/layers/pooling_layer.cpp
View File

@ -114,11 +114,15 @@ public:
Ptr<OCL4DNNPool<float> > poolOp;
#endif
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(!inputs.empty());
cv::Size inp(inputs[0]->size[3], inputs[0]->size[2]),
cv::Size inp(inputs[0].size[3], inputs[0].size[2]),
out(outputs[0].size[3], outputs[0].size[2]);
if(globalPooling)
@ -204,28 +208,29 @@ public:
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
}
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
switch (type)
{
case MAX:
CV_Assert_N(inputs.size() == 1, outputs.size() == 2);
maxPooling(*inputs[0], outputs[0], outputs[1]);
maxPooling(inputs[0], outputs[0], outputs[1]);
break;
case AVE:
CV_Assert_N(inputs.size() == 1, outputs.size() == 1);
avePooling(*inputs[0], outputs[0]);
avePooling(inputs[0], outputs[0]);
break;
case ROI: case PSROI:
CV_Assert_N(inputs.size() == 2, outputs.size() == 1);
roiPooling(*inputs[0], *inputs[1], outputs[0]);
roiPooling(inputs[0], inputs[1], outputs[0]);
break;
default:
CV_Error(Error::StsNotImplemented, "Not implemented");

37
modules/dnn/src/layers/prior_box_layer.cpp
View File

@ -297,15 +297,18 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
CV_CheckGT(inputs.size(), (size_t)1, "");
CV_CheckEQ(inputs[0]->dims, 4, ""); CV_CheckEQ(inputs[1]->dims, 4, "");
int layerWidth = inputs[0]->size[3];
int layerHeight = inputs[0]->size[2];
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
int imageWidth = inputs[1]->size[3];
int imageHeight = inputs[1]->size[2];
CV_CheckGT(inputs.size(), (size_t)1, "");
CV_CheckEQ(inputs[0].dims, 4, ""); CV_CheckEQ(inputs[1].dims, 4, "");
int layerWidth = inputs[0].size[3];
int layerHeight = inputs[0].size[2];
int imageWidth = inputs[1].size[3];
int imageHeight = inputs[1].size[2];
_stepY = _stepY == 0 ? (static_cast<float>(imageHeight) / layerHeight) : _stepY;
_stepX = _stepX == 0 ? (static_cast<float>(imageWidth) / layerWidth) : _stepX;
@ -403,21 +406,23 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(inputs.size() == 2);
int _layerWidth = inputs[0]->size[3];
int _layerHeight = inputs[0]->size[2];
int _layerWidth = inputs[0].size[3];
int _layerHeight = inputs[0].size[2];
int _imageWidth = inputs[1]->size[3];
int _imageHeight = inputs[1]->size[2];
int _imageWidth = inputs[1].size[3];
int _imageHeight = inputs[1].size[2];
float* outputPtr = outputs[0].ptr<float>();
float _boxWidth, _boxHeight;

38
modules/dnn/src/layers/proposal_layer.cpp
View File

@ -137,24 +137,27 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat*> layerInputs;
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
std::vector<Mat> layerInputs;
std::vector<Mat> layerOutputs;
// Scores permute layer.
Mat scores = getObjectScores(*inputs[0]);
layerInputs.assign(1, &scores);
Mat scores = getObjectScores(inputs[0]);
layerInputs.assign(1, scores);
layerOutputs.assign(1, Mat(shape(scores.size[0], scores.size[2],
scores.size[3], scores.size[1]), CV_32FC1));
scoresPermute->finalize(layerInputs, layerOutputs);
// BBox predictions permute layer.
Mat* bboxDeltas = inputs[1];
CV_Assert(bboxDeltas->dims == 4);
const Mat& bboxDeltas = inputs[1];
CV_Assert(bboxDeltas.dims == 4);
layerInputs.assign(1, bboxDeltas);
layerOutputs.assign(1, Mat(shape(bboxDeltas->size[0], bboxDeltas->size[2],
bboxDeltas->size[3], bboxDeltas->size[1]), CV_32FC1));
layerOutputs.assign(1, Mat(shape(bboxDeltas.size[0], bboxDeltas.size[2],
bboxDeltas.size[3], bboxDeltas.size[1]), CV_32FC1));
deltasPermute->finalize(layerInputs, layerOutputs);
}
@ -251,19 +254,22 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
CV_Assert(inputs.size() == 3);
CV_Assert(internals.size() == 3);
const Mat& scores = *inputs[0];
const Mat& bboxDeltas = *inputs[1];
const Mat& imInfo = *inputs[2];
const Mat& scores = inputs[0];
const Mat& bboxDeltas = inputs[1];
const Mat& imInfo = inputs[2];
Mat& priorBoxes = internals[0];
Mat& permuttedScores = internals[1];
Mat& permuttedDeltas = internals[2];

44
modules/dnn/src/layers/recurrent_layers.cpp
View File

@ -216,11 +216,14 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &input, std::vector<Mat> &output) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> input;
inputs_arr.getMatVector(input);
CV_Assert(!usePeephole && blobs.size() == 3 || usePeephole && blobs.size() == 6);
CV_Assert(input.size() == 1);
const Mat& inp0 = *input[0];
const Mat& inp0 = input[0];
Mat &Wh = blobs[0], &Wx = blobs[1];
int numOut = Wh.size[1];
@ -256,13 +259,16 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &input, std::vector<Mat> &output, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> input, output, internals;
inputs_arr.getMatVector(input);
outputs_arr.getMatVector(output);
internals_arr.getMatVector(internals);
const Mat &Wh = blobs[0];
const Mat &Wx = blobs[1];
@ -277,7 +283,7 @@ public:
dummyOnes.setTo(1.);
int numSamplesTotal = numTimeStamps*numSamples;
Mat xTs = input[0]->reshape(1, numSamplesTotal);
Mat xTs = input[0].reshape(1, numSamplesTotal);
Mat hOutTs = output[0].reshape(1, numSamplesTotal);
Mat cOutTs = produceCellOutput ? output[1].reshape(1, numSamplesTotal) : Mat();
@ -432,8 +438,11 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &input, std::vector<Mat> &output) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> input, outputs;
inputs_arr.getMatVector(input);
CV_Assert(input.size() >= 1 && input.size() <= 2);
Wxh = blobs[0];
@ -446,7 +455,7 @@ public:
numX = Wxh.cols;
numO = Who.rows;
const Mat& inp0 = *input[0];
const Mat& inp0 = input[0];
CV_Assert(inp0.dims >= 2);
CV_Assert(inp0.total(2) == numX);
@ -477,15 +486,18 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &input, std::vector<Mat> &output, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> input, output, internals;
inputs_arr.getMatVector(input);
outputs_arr.getMatVector(output);
internals_arr.getMatVector(internals);
Mat xTs = input[0]->reshape(1, numSamplesTotal);
Mat xTs = input[0].reshape(1, numSamplesTotal);
Mat oTs = output[0].reshape(1, numSamplesTotal);
Mat hTs = produceH ? output[1].reshape(1, numSamplesTotal) : Mat();
Mat hCurr = internals[0];

21
modules/dnn/src/layers/region_layer.cpp
View File

@ -190,13 +190,16 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
CV_Assert(inputs.size() >= 1);
CV_Assert(outputs.size() == 1);
@ -206,14 +209,14 @@ public:
for (size_t ii = 0; ii < outputs.size(); ii++)
{
Mat &inpBlob = *inputs[ii];
Mat &inpBlob = inputs[ii];
Mat &outBlob = outputs[ii];
int rows = inpBlob.size[1];
int cols = inpBlob.size[2];
CV_Assert(inputs.size() < 2 || inputs[1]->dims == 4);
int hNorm = inputs.size() > 1 ? inputs[1]->size[2] : rows;
int wNorm = inputs.size() > 1 ? inputs[1]->size[3] : cols;
CV_Assert(inputs.size() < 2 || inputs[1].dims == 4);
int hNorm = inputs.size() > 1 ? inputs[1].size[2] : rows;
int wNorm = inputs.size() > 1 ? inputs[1].size[3] : cols;
const float *srcData = inpBlob.ptr<float>();
float *dstData = outBlob.ptr<float>();

14
modules/dnn/src/layers/reorg_layer.cpp
View File

@ -139,17 +139,19 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
for (size_t i = 0; i < inputs.size(); i++)
{
Mat srcBlob = *inputs[i];
Mat srcBlob = inputs[i];
MatShape inputShape = shape(srcBlob), outShape = shape(outputs[i]);
float *dstData = outputs[0].ptr<float>();
const float *srcData = srcBlob.ptr<float>();

15
modules/dnn/src/layers/reshape_layer.cpp
View File

@ -237,17 +237,18 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
for (size_t i = 0; i < outputs.size(); i++)
{
Mat srcBlob = *inputs[i];
Mat srcBlob = inputs[i];
if (outputs[i].data != srcBlob.data)
srcBlob.reshape(1, shape(outputs[i])).copyTo(outputs[i]);
}

45
modules/dnn/src/layers/resize_layer.cpp
View File

@ -57,22 +57,26 @@ public:
return backendId == DNN_BACKEND_OPENCV;
}
virtual void finalize(const std::vector<Mat*>& inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
if (!outWidth && !outHeight)
{
outHeight = outputs[0].size[2];
outWidth = outputs[0].size[3];
}
if (alignCorners && outHeight > 1)
scaleHeight = static_cast<float>(inputs[0]->size[2] - 1) / (outHeight - 1);
scaleHeight = static_cast<float>(inputs[0].size[2] - 1) / (outHeight - 1);
else
scaleHeight = static_cast<float>(inputs[0]->size[2]) / outHeight;
scaleHeight = static_cast<float>(inputs[0].size[2]) / outHeight;
if (alignCorners && outWidth > 1)
scaleWidth = static_cast<float>(inputs[0]->size[3] - 1) / (outWidth - 1);
scaleWidth = static_cast<float>(inputs[0].size[3] - 1) / (outWidth - 1);
else
scaleWidth = static_cast<float>(inputs[0]->size[3]) / outWidth;
scaleWidth = static_cast<float>(inputs[0].size[3]) / outWidth;
}
void forward(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays internals_arr) CV_OVERRIDE
@ -80,24 +84,27 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
if (outHeight == inputs[0]->size[2] && outWidth == inputs[0]->size[3])
if (outHeight == inputs[0].size[2] && outWidth == inputs[0].size[3])
return;
Mat& inp = *inputs[0];
Mat& inp = inputs[0];
Mat& out = outputs[0];
if (interpolation == "nearest")
{
for (size_t n = 0; n < inputs[0]->size[0]; ++n)
for (size_t n = 0; n < inputs[0].size[0]; ++n)
{
for (size_t ch = 0; ch < inputs[0]->size[1]; ++ch)
for (size_t ch = 0; ch < inputs[0].size[1]; ++ch)
{
resize(getPlane(inp, n, ch), getPlane(out, n, ch),
Size(outWidth, outHeight), 0, 0, INTER_NEAREST);
@ -203,15 +210,19 @@ public:
return backendId == DNN_BACKEND_OPENCV || backendId == DNN_BACKEND_INFERENCE_ENGINE;
}
virtual void finalize(const std::vector<Mat*>& inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
if (!outWidth && !outHeight)
{
outHeight = outputs[0].size[2];
outWidth = outputs[0].size[3];
}
int inpHeight = inputs[0]->size[2];
int inpWidth = inputs[0]->size[3];
int inpHeight = inputs[0].size[2];
int inpWidth = inputs[0].size[3];
scaleHeight = (outHeight > 1) ? (static_cast<float>(inpHeight - 1) / (outHeight - 1)) : 0.f;
scaleWidth = (outWidth > 1) ? (static_cast<float>(inpWidth - 1) / (outWidth - 1)) : 0.f;
}

21
modules/dnn/src/layers/scale_layer.cpp
View File

@ -40,8 +40,10 @@ public:
return true;
}
virtual void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
std::vector<Mat> inputs;
inputs_arr.getMatVector(inputs);
hasWeights = blobs.size() == 2 || (blobs.size() == 1 && !hasBias);
CV_Assert(inputs.size() == 2 && blobs.empty() || blobs.size() == (int)hasWeights + (int)hasBias);
}
@ -57,20 +59,23 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert_N(outputs.size() == 1, !blobs.empty() || inputs.size() == 2);
Mat &inpBlob = *inputs[0];
Mat &inpBlob = inputs[0];
Mat &outBlob = outputs[0];
// There is a mode when we multiply a first blob by a second one
// instead of trainable weights.
Mat weights = blobs.empty() ? *inputs[1] : (hasWeights ? blobs[0] : Mat());
Mat weights = blobs.empty() ? inputs[1] : (hasWeights ? blobs[0] : Mat());
Mat bias = hasBias ? blobs.back().reshape(1, 1) : Mat();
if (!weights.empty())
weights = weights.reshape(1, 1);

34
modules/dnn/src/layers/shuffle_channel_layer.cpp
View File

@ -28,17 +28,21 @@ public:
return group == 1;
}
virtual void finalize(const std::vector<Mat*>& inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
if (group != 1)
{
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
LayerParams lp;
float order[] = {0, 2, 1, 3};
lp.set("order", DictValue::arrayInt(&order[0], 4));
permute = PermuteLayer::create(lp);
Mat inp = *inputs[0];
Mat out = outputs[0];
const Mat& inp = inputs[0];
const Mat& out = outputs[0];
permuteInpShape.resize(4);
permuteInpShape[0] = inp.size[0];
@ -52,11 +56,8 @@ public:
permuteOutShape[2] = permuteInpShape[1];
permuteOutShape[3] = permuteInpShape[3];
inp = inp.reshape(1, permuteInpShape);
out = out.reshape(1, permuteOutShape);
std::vector<Mat*> permuteInputs(1, &inp);
std::vector<Mat> permuteOutputs(1, out);
std::vector<Mat> permuteInputs(1, inp.reshape(1, permuteInpShape));
std::vector<Mat> permuteOutputs(1, out.reshape(1, permuteOutShape));
permute->finalize(permuteInputs, permuteOutputs);
}
}
@ -66,15 +67,18 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
Mat inp = *inputs[0];
Mat inp = inputs[0];
Mat out = outputs[0];
if (inp.data != out.data)
{
@ -82,7 +86,7 @@ public:
{
inp = inp.reshape(1, permuteInpShape);
out = out.reshape(1, permuteOutShape);
std::vector<Mat*> permuteInputs(1, &inp);
std::vector<Mat> permuteInputs(1, inp);
std::vector<Mat> permuteOutputs(1, out);
permute->forward(permuteInputs, permuteOutputs, internals);
}

22
modules/dnn/src/layers/slice_layer.cpp
View File

@ -144,10 +144,14 @@ public:
return false;
}
void finalize(const std::vector<Mat*> &inputs, std::vector<Mat> &outputs) CV_OVERRIDE
void finalize(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
CV_Assert(inputs.size() == 1);
const MatSize& inpShape = inputs[0]->size;
const MatSize& inpShape = inputs[0].size;
if (sliceRanges.empty())
{
@ -239,15 +243,17 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
const Mat& inpMat = *inputs[0];
const Mat& inpMat = inputs[0];
CV_Assert(outputs.size() == sliceRanges.size());
for (size_t i = 0; i < outputs.size(); i++)
{

15
modules/dnn/src/layers/softmax_layer.cpp
View File

@ -191,15 +191,18 @@ public:
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs, internals;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
const Mat &src = *inputs[0];
const Mat &src = inputs[0];
Mat &dst = outputs[0];
int axis = clamp(axisRaw, src.dims);

17
modules/dnn/src/layers/split_layer.cpp
View File

@ -83,18 +83,19 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
Layer::forward_fallback(inputs_arr, outputs_arr, internals_arr);
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
for (size_t i = 0; i < outputs.size(); i++)
{
CV_Assert(inputs[0]->total() == outputs[i].total());
inputs[0]->copyTo(outputs[i]);
CV_Assert(inputs[0].total() == outputs[i].total());
inputs[0].copyTo(outputs[i]);
}
}
};

6
modules/dnn/src/op_inf_engine.cpp
View File

@ -551,12 +551,6 @@ bool InfEngineBackendLayer::supportBackend(int backendId)
backendId == DNN_BACKEND_INFERENCE_ENGINE && haveInfEngine();
}
void InfEngineBackendLayer::forward(std::vector<Mat*> &input, std::vector<Mat> &output,
std::vector<Mat> &internals)
{
CV_Error(Error::StsError, "Choose Inference Engine as a preferable backend.");
}
void InfEngineBackendLayer::forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs,
OutputArrayOfArrays internals)
{

3
modules/dnn/src/op_inf_engine.hpp
View File

@ -196,9 +196,6 @@ public:
std::vector<MatShape> &outputs,
std::vector<MatShape> &internals) const CV_OVERRIDE;
virtual void forward(std::vector<Mat*> &input, std::vector<Mat> &output,
std::vector<Mat> &internals) CV_OVERRIDE;
virtual void forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs,
OutputArrayOfArrays internals) CV_OVERRIDE;

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

@ -61,16 +61,13 @@ static String _tf(TString filename)
void runLayer(Ptr<Layer> layer, std::vector<Mat> &inpBlobs, std::vector<Mat> &outBlobs)
{
size_t ninputs = inpBlobs.size();
std::vector<Mat> inp_(ninputs);
std::vector<Mat*> inp(ninputs);
std::vector<Mat> outp, intp;
std::vector<Mat> inp(ninputs), outp, intp;
std::vector<MatShape> inputs, outputs, internals;
for (size_t i = 0; i < ninputs; i++)
{
inp_[i] = inpBlobs[i].clone();
inp[i] = &inp_[i];
inputs.push_back(shape(inp_[i]));
inp[i] = inpBlobs[i].clone();
inputs.push_back(shape(inp[i]));
}
layer->getMemoryShapes(inputs, 0, outputs, internals);
@ -1052,8 +1049,6 @@ public:
return backendId == DNN_BACKEND_OPENCV;
}
virtual void forward(std::vector<cv::Mat*> &inputs, std::vector<cv::Mat> &outputs, std::vector<cv::Mat> &internals) CV_OVERRIDE {}
virtual void forward(cv::InputArrayOfArrays inputs, cv::OutputArrayOfArrays outputs, cv::OutputArrayOfArrays internals) CV_OVERRIDE {}
};
@ -1151,8 +1146,11 @@ public:
return false;
}
virtual void finalize(const std::vector<Mat*>& inputs, std::vector<Mat> &outputs) CV_OVERRIDE
virtual void finalize(InputArrayOfArrays, OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<Mat> outputs;
outputs_arr.getMatVector(outputs);
if (!outWidth && !outHeight)
{
outHeight = outputs[0].size[2];
@ -1161,9 +1159,22 @@ public:
}
// Implementation of this custom layer is based on https://github.com/cdmh/deeplab-public/blob/master/src/caffe/layers/interp_layer.cpp
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat>& internals) CV_OVERRIDE
void forward(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays internals_arr) CV_OVERRIDE
{
Mat& inp = *inputs[0];
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
if (inputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
Mat& inp = inputs[0];
Mat& out = outputs[0];
const float* inpData = (float*)inp.data;
float* outData = (float*)out.data;

140
modules/dnn/test/test_misc.cpp
View File

@ -6,7 +6,8 @@
// Third party copyrights are property of their respective owners.
#include "test_precomp.hpp"
#include <opencv2/core/ocl.hpp>
#include <opencv2/core/opencl/ocl_defs.hpp>
#include <opencv2/dnn/layer.details.hpp> // CV_DNN_REGISTER_LAYER_CLASS
namespace opencv_test { namespace {
@ -87,9 +88,13 @@ public:
return Ptr<Layer>(new FirstCustomLayer(params));
}
virtual void forward(InputArrayOfArrays, OutputArrayOfArrays, OutputArrayOfArrays) CV_OVERRIDE {}
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat>& internals) CV_OVERRIDE
void forward(InputArrayOfArrays, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> outputs;
outputs_arr.getMatVector(outputs);
outputs[0].setTo(1);
}
};
@ -104,9 +109,13 @@ public:
return Ptr<Layer>(new SecondCustomLayer(params));
}
virtual void forward(InputArrayOfArrays, OutputArrayOfArrays, OutputArrayOfArrays) CV_OVERRIDE {}
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat>& internals) CV_OVERRIDE
void forward(InputArrayOfArrays, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> outputs;
outputs_arr.getMatVector(outputs);
outputs[0].setTo(2);
}
};
@ -178,4 +187,125 @@ INSTANTIATE_TEST_CASE_P(/**/, setInput, Combine(
dnnBackendsAndTargets()
));
class CustomLayerWithDeprecatedForward CV_FINAL : public Layer
{
public:
CustomLayerWithDeprecatedForward(const LayerParams &params) : Layer(params) {}
static Ptr<Layer> create(LayerParams& params)
{
return Ptr<Layer>(new CustomLayerWithDeprecatedForward(params));
}
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_Assert_N(inputs[0]->depth() == CV_32F, outputs[0].depth() == CV_32F);
cv::add(*inputs[0], 0.5f, outputs[0]);
}
};
class CustomLayerWithDeprecatedForwardAndFallback CV_FINAL : public Layer
{
public:
CustomLayerWithDeprecatedForwardAndFallback(const LayerParams &params) : Layer(params) {}
static Ptr<Layer> create(LayerParams& params)
{
return Ptr<Layer>(new CustomLayerWithDeprecatedForwardAndFallback(params));
}
void forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs, OutputArrayOfArrays internals) CV_OVERRIDE
{
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
CV_OCL_RUN(preferableTarget == DNN_TARGET_OPENCL || preferableTarget == DNN_TARGET_OPENCL_FP16,
forward_ocl(inputs, outputs, internals));
Layer::forward_fallback(inputs, outputs, internals);
}
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
{
CV_Assert_N(inputs[0]->depth() == CV_32F, outputs[0].depth() == CV_32F);
cv::add(*inputs[0], 0.5f, outputs[0]);
}
#ifdef HAVE_OPENCL
bool forward_ocl(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays internals_arr)
{
if (inputs_arr.depth() != CV_32F)
return false;
std::vector<UMat> inputs;
std::vector<UMat> outputs;
inputs_arr.getUMatVector(inputs);
outputs_arr.getUMatVector(outputs);
cv::add(inputs[0], 0.5f, outputs[0]);
return true;
}
#endif
};
typedef testing::TestWithParam<tuple<Backend, Target> > DeprecatedForward;
TEST_P(DeprecatedForward, CustomLayer)
{
const int backend = get<0>(GetParam());
const int target = get<1>(GetParam());
Mat inp(5, 5, CV_32FC1);
randu(inp, -1.0f, 1.0f);
inp = blobFromImage(inp);
CV_DNN_REGISTER_LAYER_CLASS(CustomType, CustomLayerWithDeprecatedForward);
try
{
LayerParams lp;
Net net;
net.addLayerToPrev("testLayer", "CustomType", lp);
net.setPreferableBackend(backend);
net.setPreferableTarget(target);
net.setInput(inp);
Mat out = net.forward();
normAssert(out, inp + 0.5f, "", 2e-4, 7e-4);
}
catch (...)
{
LayerFactory::unregisterLayer("CustomType");
throw;
}
LayerFactory::unregisterLayer("CustomType");
}
TEST_P(DeprecatedForward, CustomLayerWithFallback)
{
const int backend = get<0>(GetParam());
const int target = get<1>(GetParam());
Mat inp(5, 5, CV_32FC1);
randu(inp, -1.0f, 1.0f);
inp = blobFromImage(inp);
CV_DNN_REGISTER_LAYER_CLASS(CustomType, CustomLayerWithDeprecatedForwardAndFallback);
try
{
LayerParams lp;
Net net;
net.addLayerToPrev("testLayer", "CustomType", lp);
net.setPreferableBackend(backend);
net.setPreferableTarget(target);
net.setInput(inp);
Mat out = net.forward();
normAssert(out, inp + 0.5f, "", 2e-4, 7e-4);
}
catch (...)
{
LayerFactory::unregisterLayer("CustomType");
throw;
}
LayerFactory::unregisterLayer("CustomType");
}
INSTANTIATE_TEST_CASE_P(/**/, DeprecatedForward, dnnBackendsAndTargets());
}} // namespace

15
modules/dnn/test/test_torch_importer.cpp
View File

@ -411,15 +411,22 @@ public:
return false;
}
virtual void forward(std::vector<Mat*> &inputs, std::vector<Mat> &outputs, std::vector<Mat> &internals) CV_OVERRIDE
void forward(InputArrayOfArrays inputs_arr, OutputArrayOfArrays outputs_arr, OutputArrayOfArrays) CV_OVERRIDE
{
Mat& inp = *inputs[0];
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
std::vector<Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
Mat& inp = inputs[0];
Mat& out = outputs[0];
const int outHeight = out.size[2];
const int outWidth = out.size[3];
for (size_t n = 0; n < inputs[0]->size[0]; ++n)
for (size_t n = 0; n < inp.size[0]; ++n)
{
for (size_t ch = 0; ch < inputs[0]->size[1]; ++ch)
for (size_t ch = 0; ch < inp.size[1]; ++ch)
{
resize(getPlane(inp, n, ch), getPlane(out, n, ch),
Size(outWidth, outHeight), 0, 0, INTER_NEAREST);

6
modules/java/common.cmake
View File

@ -5,6 +5,12 @@ else()
ocv_update(OPENCV_JAVA_LIB_NAME_SUFFIX "${OPENCV_VERSION_MAJOR}${OPENCV_VERSION_MINOR}${OPENCV_VERSION_PATCH}")
endif()
if(MSVC)
ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4996)
else()
ocv_warnings_disable(CMAKE_CXX_FLAGS -Wdeprecated-declarations)
endif()
# get list of modules to wrap
# message(STATUS "Wrapped in java:")
set(OPENCV_JAVA_MODULES)

54
samples/dnn/custom_layers.hpp
View File

@ -35,10 +35,23 @@ public:
}
// Implementation of this custom layer is based on https://github.com/cdmh/deeplab-public/blob/master/src/caffe/layers/interp_layer.cpp
virtual void forward(std::vector<cv::Mat*> &inputs, std::vector<cv::Mat> &outputs, std::vector<cv::Mat> &internals) CV_OVERRIDE
virtual void forward(cv::InputArrayOfArrays inputs_arr,
cv::OutputArrayOfArrays outputs_arr,
cv::OutputArrayOfArrays internals_arr) CV_OVERRIDE
{
CV_UNUSED(internals);
cv::Mat& inp = *inputs[0];
if (inputs_arr.depth() == CV_16S)
{
// In case of DNN_TARGET_OPENCL_FP16 target the following method
// converts data from FP16 to FP32 and calls this forward again.
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
std::vector<cv::Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
cv::Mat& inp = inputs[0];
cv::Mat& out = outputs[0];
const float* inpData = (float*)inp.data;
float* outData = (float*)out.data;
@ -78,8 +91,6 @@ public:
}
}
virtual void forward(cv::InputArrayOfArrays, cv::OutputArrayOfArrays, cv::OutputArrayOfArrays) CV_OVERRIDE {}
private:
int outWidth, outHeight;
};
@ -134,8 +145,10 @@ public:
return false;
}
virtual void finalize(const std::vector<cv::Mat*>&, std::vector<cv::Mat> &outputs) CV_OVERRIDE
virtual void finalize(cv::InputArrayOfArrays, cv::OutputArrayOfArrays outputs_arr) CV_OVERRIDE
{
std::vector<cv::Mat> outputs;
outputs_arr.getMatVector(outputs);
if (!outWidth && !outHeight)
{
outHeight = outputs[0].size[2];
@ -145,9 +158,23 @@ public:
// This implementation is based on a reference implementation from
// https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/lite/kernels/internal/reference/reference_ops.h
virtual void forward(std::vector<cv::Mat*> &inputs, std::vector<cv::Mat> &outputs, std::vector<cv::Mat> &) CV_OVERRIDE
virtual void forward(cv::InputArrayOfArrays inputs_arr,
cv::OutputArrayOfArrays outputs_arr,
cv::OutputArrayOfArrays internals_arr) CV_OVERRIDE
{
cv::Mat& inp = *inputs[0];
if (inputs_arr.depth() == CV_16S)
{
// In case of DNN_TARGET_OPENCL_FP16 target the following method
// converts data from FP16 to FP32 and calls this forward again.
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
std::vector<cv::Mat> inputs, outputs;
inputs_arr.getMatVector(inputs);
outputs_arr.getMatVector(outputs);
cv::Mat& inp = inputs[0];
cv::Mat& out = outputs[0];
const float* inpData = (float*)inp.data;
float* outData = (float*)out.data;
@ -185,8 +212,6 @@ public:
}
}
virtual void forward(cv::InputArrayOfArrays, cv::OutputArrayOfArrays, cv::OutputArrayOfArrays) CV_OVERRIDE {}
private:
static inline int offset(const cv::MatSize& size, int c, int x, int y, int b)
{
@ -221,14 +246,15 @@ public:
//! [MyLayer::getMemoryShapes]
//! [MyLayer::forward]
virtual void forward(std::vector<cv::Mat*> &inputs, std::vector<cv::Mat> &outputs, std::vector<cv::Mat> &internals) CV_OVERRIDE;
virtual void forward(cv::InputArrayOfArrays inputs,
cv::OutputArrayOfArrays outputs,
cv::OutputArrayOfArrays internals) CV_OVERRIDE;
//! [MyLayer::forward]
//! [MyLayer::finalize]
virtual void finalize(const std::vector<cv::Mat*> &inputs, std::vector<cv::Mat> &outputs) CV_OVERRIDE;
virtual void finalize(cv::InputArrayOfArrays inputs,
cv::OutputArrayOfArrays outputs) CV_OVERRIDE;
//! [MyLayer::finalize]
virtual void forward(cv::InputArrayOfArrays inputs, cv::OutputArrayOfArrays outputs, cv::OutputArrayOfArrays internals) CV_OVERRIDE;
};
//! [A custom layer interface]