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

cmake/OpenCVDetectInferenceEngine.cmake

@@ -129,9 +129,9 @@ endif()
 
 if(INF_ENGINE_TARGET)
   if(NOT INF_ENGINE_RELEASE)
-    message(WARNING "InferenceEngine version has not been set, 2020.2 will be used by default. Set INF_ENGINE_RELEASE variable if you experience build errors.")
+    message(WARNING "InferenceEngine version has not been set, 2020.3 will be used by default. Set INF_ENGINE_RELEASE variable if you experience build errors.")
   endif()
-  set(INF_ENGINE_RELEASE "2020020000" CACHE STRING "Force IE version, should be in form YYYYAABBCC (e.g. 2020.1.0.2 -> 2020010002)")
+  set(INF_ENGINE_RELEASE "2020030000" CACHE STRING "Force IE version, should be in form YYYYAABBCC (e.g. 2020.1.0.2 -> 2020010002)")
   set_target_properties(${INF_ENGINE_TARGET} PROPERTIES
     INTERFACE_COMPILE_DEFINITIONS "HAVE_INF_ENGINE=1;INF_ENGINE_RELEASE=${INF_ENGINE_RELEASE}"
   )

cmake/OpenCVFindMKL.cmake

@@ -79,9 +79,10 @@ get_mkl_version(${MKL_INCLUDE_DIRS}/mkl_version.h)
 
 #determine arch
 if(CMAKE_CXX_SIZEOF_DATA_PTR EQUAL 8)
-    set(MKL_X64 1)
-    set(MKL_ARCH "intel64")
-
+    set(MKL_ARCH_LIST "intel64")
+    if(MSVC)
+        list(APPEND MKL_ARCH_LIST "win-x64")
+    endif()
     include(CheckTypeSize)
     CHECK_TYPE_SIZE(int _sizeof_int)
     if (_sizeof_int EQUAL 4)
@@ -90,14 +91,19 @@ if(CMAKE_CXX_SIZEOF_DATA_PTR EQUAL 8)
         set(MKL_ARCH_SUFFIX "ilp64")
     endif()
 else()
-    set(MKL_ARCH "ia32")
+    set(MKL_ARCH_LIST "ia32")
     set(MKL_ARCH_SUFFIX "c")
 endif()
 
 if(MKL_VERSION_STR VERSION_GREATER "11.3.0" OR MKL_VERSION_STR VERSION_EQUAL "11.3.0")
     set(mkl_lib_find_paths
-        ${MKL_ROOT_DIR}/lib
-        ${MKL_ROOT_DIR}/lib/${MKL_ARCH} ${MKL_ROOT_DIR}/../tbb/lib/${MKL_ARCH})
+        ${MKL_ROOT_DIR}/lib)
+    foreach(MKL_ARCH ${MKL_ARCH_LIST})
+      list(APPEND mkl_lib_find_paths
+        ${MKL_ROOT_DIR}/lib/${MKL_ARCH}
+        ${MKL_ROOT_DIR}/../tbb/lib/${MKL_ARCH}
+        ${MKL_ROOT_DIR}/${MKL_ARCH})
+    endforeach()
 
     set(mkl_lib_list "mkl_intel_${MKL_ARCH_SUFFIX}")
 
@@ -121,7 +127,7 @@ endif()
 
 set(MKL_LIBRARIES "")
 foreach(lib ${mkl_lib_list})
-    find_library(${lib} ${lib} ${mkl_lib_find_paths})
+    find_library(${lib} NAMES ${lib} ${lib}_dll HINTS ${mkl_lib_find_paths})
     mark_as_advanced(${lib})
     if(NOT ${lib})
         mkl_fail()

doc/tutorials/introduction/macos_install/macos_install.markdown

@@ -30,7 +30,7 @@ Installing CMake
 
 -# Install the dmg package and launch it from Applications. That will give you the UI app of CMake
 
--# From the CMake app window, choose menu Tools --> Install For Command Line Use.
+-# From the CMake app window, choose menu Tools --> How to Install For Command Line Use. Then, follow the instructions from the pop-up there.
 
 -# Install folder will be /usr/bin/ by default, submit it by choosing Install command line links.
 
@@ -66,7 +66,7 @@ git clone https://github.com/opencv/opencv_contrib.git
 Building OpenCV from Source Using CMake
 ---------------------------------------
 
--#  Create a temporary directory, which we denote as `<cmake_build_dir>`, where you want to put
+-#  Create a temporary directory, which we denote as `build_opencv`, where you want to put
     the generated Makefiles, project files as well the object files and output binaries and enter
     there.
 
@@ -87,8 +87,8 @@ Building OpenCV from Source Using CMake
 
     or cmake-gui
 
-    -   set full path to OpenCV source code, e.g. `/home/user/opencv`
-    -   set full path to `<cmake_build_dir>`, e.g. `/home/user/build_opencv`
+    -   set the OpenCV source code path to, e.g. `/home/user/opencv`
+    -   set the binary build path to your CMake build directory, e.g. `/home/user/build_opencv`
     -   set optional parameters
     -   run: "Configure"
     -   run: "Generate"

modules/core/include/opencv2/core/eigen.hpp

@@ -66,10 +66,18 @@
 namespace cv
 {
 
-//! @addtogroup core_eigen
+/** @addtogroup core_eigen
+These functions are provided for OpenCV-Eigen interoperability. They convert `Mat`
+objects to corresponding `Eigen::Matrix` objects and vice-versa. Consult the [Eigen
+documentation](https://eigen.tuxfamily.org/dox/group__TutorialMatrixClass.html) for
+information about the `Matrix` template type.
+
+@note Using these functions requires the `Eigen/Dense` or similar header to be
+included before this header.
+*/
 //! @{
 
-#ifdef OPENCV_EIGEN_TENSOR_SUPPORT
+#if defined(OPENCV_EIGEN_TENSOR_SUPPORT) || defined(CV_DOXYGEN)
 /** @brief Converts an Eigen::Tensor to a cv::Mat.
 
 The method converts an Eigen::Tensor with shape (H x W x C) to a cv::Mat where:

modules/dnn/src/dnn.cpp

@@ -2248,7 +2248,7 @@ struct Net::Impl : public detail::NetImplBase
 
                     auto ieInpNode = inputNodes[i].dynamicCast<InfEngineNgraphNode>();
                     CV_Assert(oid < ieInpNode->node->get_output_size());
-#if INF_ENGINE_VER_MAJOR_GT(2020030000)
+#if INF_ENGINE_VER_MAJOR_GT(INF_ENGINE_RELEASE_2020_3)
                     inputNodes[i] = Ptr<BackendNode>(new InfEngineNgraphNode(ieInpNode->node->get_output_as_single_output_node(oid)));
 #else
                     inputNodes[i] = Ptr<BackendNode>(new InfEngineNgraphNode(ieInpNode->node->get_output_as_single_output_node(oid, false)));

modules/dnn/src/ie_ngraph.cpp

@@ -82,7 +82,7 @@ public:
         return type_info;
     }
 
-#if INF_ENGINE_VER_MAJOR_GT(2020020000)
+#if INF_ENGINE_VER_MAJOR_GE(INF_ENGINE_RELEASE_2020_3)
     NgraphCustomOp(const ngraph::OutputVector& inputs,
 #else
     NgraphCustomOp(const ngraph::NodeVector& inputs,
@@ -112,7 +112,7 @@ public:
 
     std::shared_ptr<ngraph::Node> copy_with_new_args(const ngraph::NodeVector& new_args) const override
     {
-#if INF_ENGINE_VER_MAJOR_GT(2020020000)
+#if INF_ENGINE_VER_MAJOR_GE(INF_ENGINE_RELEASE_2020_3)
         return std::make_shared<NgraphCustomOp>(ngraph::as_output_vector(new_args), params);
 #else
         return std::make_shared<NgraphCustomOp>(new_args, params);
@@ -239,7 +239,9 @@ private:
 class InfEngineNgraphExtension : public InferenceEngine::IExtension
 {
 public:
+#if INF_ENGINE_VER_MAJOR_LT(INF_ENGINE_RELEASE_2020_2)
     virtual void SetLogCallback(InferenceEngine::IErrorListener&) noexcept {}
+#endif
     virtual void Unload() noexcept {}
     virtual void Release() noexcept {}
     virtual void GetVersion(const InferenceEngine::Version*&) const noexcept {}
@@ -283,7 +285,7 @@ InfEngineNgraphNode::InfEngineNgraphNode(const std::vector<Ptr<BackendNode> >& n
         {"internals", shapesToStr(internals)}
     };
 
-#if INF_ENGINE_VER_MAJOR_GT(2020020000)
+#if INF_ENGINE_VER_MAJOR_GE(INF_ENGINE_RELEASE_2020_3)
     ngraph::OutputVector inp_nodes;
 #else
     ngraph::NodeVector inp_nodes;

modules/dnn/src/op_inf_engine.hpp

@@ -25,10 +25,11 @@
 #define INF_ENGINE_RELEASE_2019R3 2019030000
 #define INF_ENGINE_RELEASE_2020_1 2020010000
 #define INF_ENGINE_RELEASE_2020_2 2020020000
+#define INF_ENGINE_RELEASE_2020_3 2020030000
 
 #ifndef INF_ENGINE_RELEASE
-#warning("IE version have not been provided via command-line. Using 2020.2 by default")
-#define INF_ENGINE_RELEASE INF_ENGINE_RELEASE_2020_2
+#warning("IE version have not been provided via command-line. Using 2020.3 by default")
+#define INF_ENGINE_RELEASE INF_ENGINE_RELEASE_2020_3
 #endif
 
 #define INF_ENGINE_VER_MAJOR_GT(ver) (((INF_ENGINE_RELEASE) / 10000) > ((ver) / 10000))
@@ -226,7 +227,9 @@ private:
 class InfEngineExtension : public InferenceEngine::IExtension
 {
 public:
+#if INF_ENGINE_VER_MAJOR_LT(INF_ENGINE_RELEASE_2020_2)
     virtual void SetLogCallback(InferenceEngine::IErrorListener&) noexcept {}
+#endif
     virtual void Unload() noexcept {}
     virtual void Release() noexcept {}
     virtual void GetVersion(const InferenceEngine::Version*&) const noexcept {}

samples/dnn/text_detection.py

@@ -1,25 +1,81 @@
+'''
+    Text detection model: https://github.com/argman/EAST
+    Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1
+    Text recognition model taken from here: https://github.com/meijieru/crnn.pytorch
+    How to convert from pb to onnx:
+    Using classes from here: https://github.com/meijieru/crnn.pytorch/blob/master/models/crnn.py
+    import torch
+    import models.crnn as CRNN
+    model = CRNN(32, 1, 37, 256)
+    model.load_state_dict(torch.load('crnn.pth'))
+    dummy_input = torch.randn(1, 1, 32, 100)
+    torch.onnx.export(model, dummy_input, "crnn.onnx", verbose=True)
+'''
+
+
 # Import required modules
+import numpy as np
 import cv2 as cv
 import math
 import argparse
 
 ############ Add argument parser for command line arguments ############
-parser = argparse.ArgumentParser(description='Use this script to run TensorFlow implementation (https://github.com/argman/EAST) of EAST: An Efficient and Accurate Scene Text Detector (https://arxiv.org/abs/1704.03155v2)')
-parser.add_argument('--input', help='Path to input image or video file. Skip this argument to capture frames from a camera.')
-parser.add_argument('--model', required=True,
-                    help='Path to a binary .pb file of model contains trained weights.')
+parser = argparse.ArgumentParser(
+    description="Use this script to run TensorFlow implementation (https://github.com/argman/EAST) of "
+                "EAST: An Efficient and Accurate Scene Text Detector (https://arxiv.org/abs/1704.03155v2)"
+                "The OCR model can be obtained from converting the pretrained CRNN model to .onnx format from the github repository https://github.com/meijieru/crnn.pytorch")
+parser.add_argument('--input',
+                    help='Path to input image or video file. Skip this argument to capture frames from a camera.')
+parser.add_argument('--model', '-m', required=True,
+                    help='Path to a binary .pb file contains trained detector network.')
+parser.add_argument('--ocr', default="crnn.onnx",
+                    help="Path to a binary .pb or .onnx file contains trained recognition network", )
 parser.add_argument('--width', type=int, default=320,
                     help='Preprocess input image by resizing to a specific width. It should be multiple by 32.')
-parser.add_argument('--height',type=int, default=320,
+parser.add_argument('--height', type=int, default=320,
                     help='Preprocess input image by resizing to a specific height. It should be multiple by 32.')
-parser.add_argument('--thr',type=float, default=0.5,
+parser.add_argument('--thr', type=float, default=0.5,
                     help='Confidence threshold.')
-parser.add_argument('--nms',type=float, default=0.4,
+parser.add_argument('--nms', type=float, default=0.4,
                     help='Non-maximum suppression threshold.')
 args = parser.parse_args()
 
+
 ############ Utility functions ############
-def decode(scores, geometry, scoreThresh):
+
+def fourPointsTransform(frame, vertices):
+    vertices = np.asarray(vertices)
+    outputSize = (100, 32)
+    targetVertices = np.array([
+        [0, outputSize[1] - 1],
+        [0, 0],
+        [outputSize[0] - 1, 0],
+        [outputSize[0] - 1, outputSize[1] - 1]], dtype="float32")
+
+    rotationMatrix = cv.getPerspectiveTransform(vertices, targetVertices)
+    result = cv.warpPerspective(frame, rotationMatrix, outputSize)
+    return result
+
+
+def decodeText(scores):
+    text = ""
+    alphabet = "0123456789abcdefghijklmnopqrstuvwxyz"
+    for i in range(scores.shape[0]):
+        c = np.argmax(scores[i][0])
+        if c != 0:
+            text += alphabet[c - 1]
+        else:
+            text += '-'
+
+    # adjacent same letters as well as background text must be removed to get the final output
+    char_list = []
+    for i in range(len(text)):
+        if text[i] != '-' and (not (i > 0 and text[i] == text[i - 1])):
+            char_list.append(text[i])
+    return ''.join(char_list)
+
+
+def decodeBoundingBoxes(scores, geometry, scoreThresh):
     detections = []
     confidences = []
 
@@ -47,7 +103,7 @@ def decode(scores, geometry, scoreThresh):
             score = scoresData[x]
 
             # If score is lower than threshold score, move to next x
-            if(score < scoreThresh):
+            if (score < scoreThresh):
                 continue
 
             # Calculate offset
@@ -66,24 +122,27 @@ def decode(scores, geometry, scoreThresh):
 
             # Find points for rectangle
             p1 = (-sinA * h + offset[0], -cosA * h + offset[1])
-            p3 = (-cosA * w + offset[0],  sinA * w + offset[1])
-            center = (0.5*(p1[0]+p3[0]), 0.5*(p1[1]+p3[1]))
-            detections.append((center, (w,h), -1*angle * 180.0 / math.pi))
+            p3 = (-cosA * w + offset[0], sinA * w + offset[1])
+            center = (0.5 * (p1[0] + p3[0]), 0.5 * (p1[1] + p3[1]))
+            detections.append((center, (w, h), -1 * angle * 180.0 / math.pi))
             confidences.append(float(score))
 
     # Return detections and confidences
     return [detections, confidences]
 
+
 def main():
     # Read and store arguments
     confThreshold = args.thr
     nmsThreshold = args.nms
     inpWidth = args.width
     inpHeight = args.height
-    model = args.model
+    modelDetector = args.model
+    modelRecognition = args.ocr
 
     # Load network
-    net = cv.dnn.readNet(model)
+    detector = cv.dnn.readNet(modelDetector)
+    recognizer = cv.dnn.readNet(modelRecognition)
 
     # Create a new named window
     kWinName = "EAST: An Efficient and Accurate Scene Text Detector"
@@ -95,6 +154,7 @@ def main():
     # Open a video file or an image file or a camera stream
     cap = cv.VideoCapture(args.input if args.input else 0)
 
+    tickmeter = cv.TickMeter()
     while cv.waitKey(1) < 0:
         # Read frame
         hasFrame, frame = cap.read()
@@ -111,19 +171,20 @@ def main():
         # Create a 4D blob from frame.
         blob = cv.dnn.blobFromImage(frame, 1.0, (inpWidth, inpHeight), (123.68, 116.78, 103.94), True, False)
 
-        # Run the model
-        net.setInput(blob)
-        outs = net.forward(outNames)
-        t, _ = net.getPerfProfile()
-        label = 'Inference time: %.2f ms' % (t * 1000.0 / cv.getTickFrequency())
+        # Run the detection model
+        detector.setInput(blob)
+
+        tickmeter.start()
+        outs = detector.forward(outNames)
+        tickmeter.stop()
 
         # Get scores and geometry
         scores = outs[0]
         geometry = outs[1]
-        [boxes, confidences] = decode(scores, geometry, confThreshold)
+        [boxes, confidences] = decodeBoundingBoxes(scores, geometry, confThreshold)
 
         # Apply NMS
-        indices = cv.dnn.NMSBoxesRotated(boxes, confidences, confThreshold,nmsThreshold)
+        indices = cv.dnn.NMSBoxesRotated(boxes, confidences, confThreshold, nmsThreshold)
         for i in indices:
             # get 4 corners of the rotated rect
             vertices = cv.boxPoints(boxes[i[0]])
@@ -131,16 +192,40 @@ def main():
             for j in range(4):
                 vertices[j][0] *= rW
                 vertices[j][1] *= rH
+
+
+            # get cropped image using perspective transform
+            if modelRecognition:
+                cropped = fourPointsTransform(frame, vertices)
+                cropped = cv.cvtColor(cropped, cv.COLOR_BGR2GRAY)
+
+                # Create a 4D blob from cropped image
+                blob = cv.dnn.blobFromImage(cropped, size=(100, 32), mean=127.5, scalefactor=1 / 127.5)
+                recognizer.setInput(blob)
+
+                # Run the recognition model
+                tickmeter.start()
+                result = recognizer.forward()
+                tickmeter.stop()
+
+                # decode the result into text
+                wordRecognized = decodeText(result)
+                cv.putText(frame, wordRecognized, (int(vertices[1][0]), int(vertices[1][1])), cv.FONT_HERSHEY_SIMPLEX,
+                           0.5, (255, 0, 0))
+
             for j in range(4):
                 p1 = (vertices[j][0], vertices[j][1])
                 p2 = (vertices[(j + 1) % 4][0], vertices[(j + 1) % 4][1])
                 cv.line(frame, p1, p2, (0, 255, 0), 1)
 
         # Put efficiency information
+        label = 'Inference time: %.2f ms' % (tickmeter.getTimeMilli())
         cv.putText(frame, label, (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0))
 
         # Display the frame
-        cv.imshow(kWinName,frame)
+        cv.imshow(kWinName, frame)
+        tickmeter.reset()
+
 
 if __name__ == "__main__":
     main()