From bd7b249195aa4bbff4e559bffd62d124f19c01cf Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Sun, 5 Apr 2020 00:49:54 +0200
Subject: [PATCH 01/10] addaption of
 https://github.com/opencv/opencv/blob/master/samples/dnn/text_detection.cpp

---
 DESCRIPTION         |   2 +-
 NAMESPACE           |   1 +
 R/RcppExports.R     |   4 +
 R/text.R            |  27 +++++++
 man/opencv.Rd       |   5 +-
 src/RcppExports.cpp |  18 +++++
 src/text.cpp        | 185 ++++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 240 insertions(+), 2 deletions(-)
 create mode 100644 R/text.R
 create mode 100644 src/text.cpp

diff --git a/DESCRIPTION b/DESCRIPTION
index eb6990b..cacf82f 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -16,5 +16,5 @@ LinkingTo: Rcpp
 Imports: Rcpp, magrittr
 LazyData: true
 Encoding: UTF-8
-RoxygenNote: 6.1.1
+RoxygenNote: 7.1.0
 Roxygen: list(markdown = TRUE)
diff --git a/NAMESPACE b/NAMESPACE
index 0d9dab4..281b878 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -20,6 +20,7 @@ export(ocv_read)
 export(ocv_resize)
 export(ocv_sketch)
 export(ocv_stylize)
+export(ocv_text)
 export(ocv_video)
 export(ocv_write)
 importFrom(Rcpp,sourceCpp)
diff --git a/R/RcppExports.R b/R/RcppExports.R
index 84d20c9..f14a751 100644
--- a/R/RcppExports.R
+++ b/R/RcppExports.R
@@ -105,3 +105,7 @@ cvmat_markers <- function(ptr) {
     .Call('_opencv_cvmat_markers', PACKAGE = 'opencv', ptr)
 }
 
+text_detection <- function(input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw) {
+    .Call('_opencv_text_detection', PACKAGE = 'opencv', input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw)
+}
+
diff --git a/R/text.R b/R/text.R
new file mode 100644
index 0000000..8c12976
--- /dev/null
+++ b/R/text.R
@@ -0,0 +1,27 @@
+#' @export
+#' @rdname opencv
+#' @param image Path to input image or video file. Skip this argument to capture frames from a camera.
+#' @param width Preprocess input image by resizing to a specific width. It should be multiple by 32.
+#' @param height Preprocess input image by resizing to a specific height. It should be multiple by 32.
+#' @param thrs Confidence threshold.
+#' @param nms Non-maximum suppression threshold.
+#'
+#' @param model Path to a binary .pb file contains trained network.
+ocv_text <- function(image, thrs = 0.5, nms = 20, width = 320, height = 320,
+                     model, draw = FALSE){
+
+  if(missing(model))
+    stop("requires a model pb-file")
+
+  image <- path.expand(image)
+  model <- path.expand(model)
+
+  text_detection(input = image,
+                 confThreshold = thrs,
+                 nmsThreshold = nms,
+                 inpWidth = width,
+                 inpHeight = height,
+                 model = model,
+                 draw = draw)
+}
+
diff --git a/man/opencv.Rd b/man/opencv.Rd
index f88d0f5..3daaf0b 100644
--- a/man/opencv.Rd
+++ b/man/opencv.Rd
@@ -1,5 +1,5 @@
 % Generated by roxygen2: do not edit by hand
-% Please edit documentation in R/face.R, R/opencv.R
+% Please edit documentation in R/face.R, R/opencv.R, R/text.R
 \name{ocv_face}
 \alias{ocv_face}
 \alias{ocv_facemask}
@@ -21,6 +21,7 @@
 \alias{ocv_copyto}
 \alias{ocv_display}
 \alias{ocv_video}
+\alias{ocv_text}
 \title{OpenCV Computer Vision}
 \usage{
 ocv_face(image)
@@ -62,6 +63,8 @@ ocv_copyto(image, target, mask)
 ocv_display(image)
 
 ocv_video(filter)
+
+ocv_text(image, thrs = 0.5, nms = 20, width = 320, height = 320, model)
 }
 \arguments{
 \item{image}{a ocv image object}
diff --git a/src/RcppExports.cpp b/src/RcppExports.cpp
index 4fe270b..42c7754 100644
--- a/src/RcppExports.cpp
+++ b/src/RcppExports.cpp
@@ -296,6 +296,23 @@ BEGIN_RCPP
     return rcpp_result_gen;
 END_RCPP
 }
+// text_detection
+Rcpp::DataFrame text_detection(std::string input, float confThreshold, float nmsThreshold, int inpWidth, int inpHeight, std::string model, bool draw);
+RcppExport SEXP _opencv_text_detection(SEXP inputSEXP, SEXP confThresholdSEXP, SEXP nmsThresholdSEXP, SEXP inpWidthSEXP, SEXP inpHeightSEXP, SEXP modelSEXP, SEXP drawSEXP) {
+BEGIN_RCPP
+    Rcpp::RObject rcpp_result_gen;
+    Rcpp::RNGScope rcpp_rngScope_gen;
+    Rcpp::traits::input_parameter< std::string >::type input(inputSEXP);
+    Rcpp::traits::input_parameter< float >::type confThreshold(confThresholdSEXP);
+    Rcpp::traits::input_parameter< float >::type nmsThreshold(nmsThresholdSEXP);
+    Rcpp::traits::input_parameter< int >::type inpWidth(inpWidthSEXP);
+    Rcpp::traits::input_parameter< int >::type inpHeight(inpHeightSEXP);
+    Rcpp::traits::input_parameter< std::string >::type model(modelSEXP);
+    Rcpp::traits::input_parameter< bool >::type draw(drawSEXP);
+    rcpp_result_gen = Rcpp::wrap(text_detection(input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw));
+    return rcpp_result_gen;
+END_RCPP
+}
 
 static const R_CallMethodDef CallEntries[] = {
     {"_opencv_cvmat_destroy", (DL_FUNC) &_opencv_cvmat_destroy, 1},
@@ -324,6 +341,7 @@ static const R_CallMethodDef CallEntries[] = {
     {"_opencv_cvmat_edges", (DL_FUNC) &_opencv_cvmat_edges, 1},
     {"_opencv_cvmat_hog", (DL_FUNC) &_opencv_cvmat_hog, 1},
     {"_opencv_cvmat_markers", (DL_FUNC) &_opencv_cvmat_markers, 1},
+    {"_opencv_text_detection", (DL_FUNC) &_opencv_text_detection, 7},
     {NULL, NULL, 0}
 };
 
diff --git a/src/text.cpp b/src/text.cpp
new file mode 100644
index 0000000..a9c879c
--- /dev/null
+++ b/src/text.cpp
@@ -0,0 +1,185 @@
+#include "util.hpp"
+#include <opencv2/imgproc.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/dnn.hpp>
+
+using namespace cv;
+using namespace cv::dnn;
+
+void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
+            std::vector<RotatedRect>& detections,
+            std::vector<float>& confidences)
+{
+  detections.clear();
+  CV_Assert(scores.dims == 4); CV_Assert(geometry.dims == 4);
+  CV_Assert(scores.size[0] == 1); CV_Assert(geometry.size[0] == 1);
+  CV_Assert(scores.size[1] == 1); CV_Assert(geometry.size[1] == 5);
+  CV_Assert(scores.size[2] == geometry.size[2]);
+  CV_Assert(scores.size[3] == geometry.size[3]);
+
+  const int height = scores.size[2];
+  const int width = scores.size[3];
+  for (int y = 0; y < height; ++y)
+  {
+    const float* scoresData = scores.ptr<float>(0, 0, y);
+    const float* x0_data = geometry.ptr<float>(0, 0, y);
+    const float* x1_data = geometry.ptr<float>(0, 1, y);
+    const float* x2_data = geometry.ptr<float>(0, 2, y);
+    const float* x3_data = geometry.ptr<float>(0, 3, y);
+    const float* anglesData = geometry.ptr<float>(0, 4, y);
+    for (int x = 0; x < width; ++x)
+    {
+      float score = scoresData[x];
+      if (score < scoreThresh)
+        continue;
+
+      // Decode a prediction.
+      // Multiple by 4 because feature maps are 4 time less than input image.
+      float offsetX = x * 4.0f, offsetY = y * 4.0f;
+      float angle = anglesData[x];
+      float cosA = std::cos(angle);
+      float sinA = std::sin(angle);
+      float h = x0_data[x] + x2_data[x];
+      float w = x1_data[x] + x3_data[x];
+
+      Point2f offset(offsetX + cosA * x1_data[x] + sinA * x2_data[x],
+                     offsetY - sinA * x1_data[x] + cosA * x2_data[x]);
+      Point2f p1 = Point2f(-sinA * h, -cosA * h) + offset;
+      Point2f p3 = Point2f(-cosA * w, sinA * w) + offset;
+      RotatedRect r(0.5f * (p1 + p3), Size2f(w, h),
+                    -angle * 180.0f / (float)CV_PI);
+      detections.push_back(r);
+      confidences.push_back(score);
+    }
+  }
+}
+
+
+// [[Rcpp::export]]
+Rcpp::DataFrame
+  text_detection(std::string input, float confThreshold,float nmsThreshold,
+                 int inpWidth, int inpHeight, std::string model, bool draw)
+{
+  if (model.empty())
+    Rcpp::stop("No model defined");
+
+  // Load network.
+  Net net = readNet(model);
+
+  // Open a video file or an image file or a camera stream.
+  VideoCapture cap;
+  if (!input.empty())
+    cap.open(input);
+  else
+    cap.open(0);
+
+  static const std::string kWinName =
+    "EAST: An Efficient and Accurate Scene Text Detector";
+
+  if (draw) {
+    namedWindow(kWinName, WINDOW_NORMAL);
+  }
+
+  std::vector<Mat> outs;
+  std::vector<String> outNames(2);
+  outNames[0] = "feature_fusion/Conv_7/Sigmoid";
+  outNames[1] = "feature_fusion/concat_3";
+
+  Mat frame, blob;
+  std::vector<int> indices;
+
+  Rcpp::IntegerVector x1vec(indices.size());
+  Rcpp::IntegerVector y1vec(indices.size());
+  Rcpp::IntegerVector x2vec(indices.size());
+  Rcpp::IntegerVector y2vec(indices.size());
+  Rcpp::IntegerVector x3vec(indices.size());
+  Rcpp::IntegerVector y3vec(indices.size());
+  Rcpp::IntegerVector x4vec(indices.size());
+  Rcpp::IntegerVector y4vec(indices.size());
+
+  // if window is drawn, push super key to exit
+  while (waitKey(1) < 0)
+  {
+    cap >> frame;
+    if (frame.empty())
+    {
+      waitKey();
+      break;
+    }
+
+    blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
+                  Scalar(123.68, 116.78, 103.94), true, false);
+    net.setInput(blob);
+    net.forward(outs, outNames);
+
+    Mat scores = outs[0];
+    Mat geometry = outs[1];
+
+    // Decode predicted bounding boxes.
+    std::vector<RotatedRect> boxes;
+    std::vector<float> confidences;
+    decode(scores, geometry, confThreshold, boxes, confidences);
+
+    // Apply non-maximum suppression procedure.
+    NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
+
+
+    // Render detections.
+    Point2f ratio((float)frame.cols / inpWidth, (float)frame.rows / inpHeight);
+    for (size_t i = 0; i < indices.size(); ++i)
+    {
+      RotatedRect& box = boxes[indices[i]];
+
+      Point2f vertices[4];
+      box.points(vertices);
+      for (int j = 0; j < 4; ++j)
+      {
+        vertices[j].x *= ratio.x;
+        vertices[j].y *= ratio.y;
+      }
+
+      // not the fastes way
+      x1vec.push_back(vertices[0].x);
+      y1vec.push_back(vertices[0].y);
+      x2vec.push_back(vertices[1].x);
+      y2vec.push_back(vertices[1].y);
+      x3vec.push_back(vertices[2].x);
+      y3vec.push_back(vertices[2].y);
+      x4vec.push_back(vertices[3].x);
+      y4vec.push_back(vertices[3].y);
+
+      if (draw)
+        for (int j = 0; j < 4; ++j)
+          line(frame, vertices[j], vertices[(j + 1) % 4], Scalar(0, 255, 0), 1);
+    }
+
+    if (draw) {
+      // Put efficiency information.
+      std::vector<double> layersTimes;
+      double freq = getTickFrequency() / 1000;
+      double t = net.getPerfProfile(layersTimes) / freq;
+      std::string label = format("Inference time: %.2f ms", t);
+      putText(frame, label, Point(0, 15),
+              FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
+    }
+
+    if (draw)
+      imshow(kWinName, frame);
+    else
+      break;
+  }
+
+  // XPtrMat out = cvmat_xptr(box);
+  // out.attr("indices") = ;
+
+  return Rcpp::DataFrame::create(
+    Rcpp::_["x1"] = x1vec,
+    Rcpp::_["y1"] = y1vec,
+    Rcpp::_["x2"] = x2vec,
+    Rcpp::_["y2"] = y2vec,
+    Rcpp::_["x3"] = x3vec,
+    Rcpp::_["y3"] = y3vec,
+    Rcpp::_["x4"] = x4vec,
+    Rcpp::_["y4"] = y4vec
+  );
+}

From 1b356e41274512d7d1af4e7b7925cd26c8f3d68f Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Sun, 5 Apr 2020 03:39:25 +0200
Subject: [PATCH 02/10] use xptrmat

---
 R/RcppExports.R     |  4 ++--
 R/text.R            |  3 +--
 src/RcppExports.cpp |  8 ++++----
 src/text.cpp        | 48 +++++++++++++--------------------------------
 4 files changed, 21 insertions(+), 42 deletions(-)

diff --git a/R/RcppExports.R b/R/RcppExports.R
index f14a751..7897015 100644
--- a/R/RcppExports.R
+++ b/R/RcppExports.R
@@ -105,7 +105,7 @@ cvmat_markers <- function(ptr) {
     .Call('_opencv_cvmat_markers', PACKAGE = 'opencv', ptr)
 }
 
-text_detection <- function(input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw) {
-    .Call('_opencv_text_detection', PACKAGE = 'opencv', input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw)
+text_detection <- function(ptr, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw) {
+    .Call('_opencv_text_detection', PACKAGE = 'opencv', ptr, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw)
 }
 
diff --git a/R/text.R b/R/text.R
index 8c12976..3a18a74 100644
--- a/R/text.R
+++ b/R/text.R
@@ -13,10 +13,9 @@ ocv_text <- function(image, thrs = 0.5, nms = 20, width = 320, height = 320,
   if(missing(model))
     stop("requires a model pb-file")
 
-  image <- path.expand(image)
   model <- path.expand(model)
 
-  text_detection(input = image,
+  text_detection(ptr = image,
                  confThreshold = thrs,
                  nmsThreshold = nms,
                  inpWidth = width,
diff --git a/src/RcppExports.cpp b/src/RcppExports.cpp
index 42c7754..9829f59 100644
--- a/src/RcppExports.cpp
+++ b/src/RcppExports.cpp
@@ -297,19 +297,19 @@ BEGIN_RCPP
 END_RCPP
 }
 // text_detection
-Rcpp::DataFrame text_detection(std::string input, float confThreshold, float nmsThreshold, int inpWidth, int inpHeight, std::string model, bool draw);
-RcppExport SEXP _opencv_text_detection(SEXP inputSEXP, SEXP confThresholdSEXP, SEXP nmsThresholdSEXP, SEXP inpWidthSEXP, SEXP inpHeightSEXP, SEXP modelSEXP, SEXP drawSEXP) {
+XPtrMat text_detection(XPtrMat ptr, float confThreshold, float nmsThreshold, int inpWidth, int inpHeight, std::string model, bool draw);
+RcppExport SEXP _opencv_text_detection(SEXP ptrSEXP, SEXP confThresholdSEXP, SEXP nmsThresholdSEXP, SEXP inpWidthSEXP, SEXP inpHeightSEXP, SEXP modelSEXP, SEXP drawSEXP) {
 BEGIN_RCPP
     Rcpp::RObject rcpp_result_gen;
     Rcpp::RNGScope rcpp_rngScope_gen;
-    Rcpp::traits::input_parameter< std::string >::type input(inputSEXP);
+    Rcpp::traits::input_parameter< XPtrMat >::type ptr(ptrSEXP);
     Rcpp::traits::input_parameter< float >::type confThreshold(confThresholdSEXP);
     Rcpp::traits::input_parameter< float >::type nmsThreshold(nmsThresholdSEXP);
     Rcpp::traits::input_parameter< int >::type inpWidth(inpWidthSEXP);
     Rcpp::traits::input_parameter< int >::type inpHeight(inpHeightSEXP);
     Rcpp::traits::input_parameter< std::string >::type model(modelSEXP);
     Rcpp::traits::input_parameter< bool >::type draw(drawSEXP);
-    rcpp_result_gen = Rcpp::wrap(text_detection(input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw));
+    rcpp_result_gen = Rcpp::wrap(text_detection(ptr, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw));
     return rcpp_result_gen;
 END_RCPP
 }
diff --git a/src/text.cpp b/src/text.cpp
index a9c879c..0c26f4e 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -56,36 +56,25 @@ void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
 
 
 // [[Rcpp::export]]
-Rcpp::DataFrame
-  text_detection(std::string input, float confThreshold,float nmsThreshold,
+XPtrMat
+  text_detection(XPtrMat ptr, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
 {
   if (model.empty())
     Rcpp::stop("No model defined");
 
+  Mat frame = get_mat(ptr);
+
   // Load network.
   Net net = readNet(model);
 
-  // Open a video file or an image file or a camera stream.
-  VideoCapture cap;
-  if (!input.empty())
-    cap.open(input);
-  else
-    cap.open(0);
-
-  static const std::string kWinName =
-    "EAST: An Efficient and Accurate Scene Text Detector";
-
-  if (draw) {
-    namedWindow(kWinName, WINDOW_NORMAL);
-  }
-
   std::vector<Mat> outs;
   std::vector<String> outNames(2);
   outNames[0] = "feature_fusion/Conv_7/Sigmoid";
   outNames[1] = "feature_fusion/concat_3";
 
-  Mat frame, blob;
+  Mat blob;
+  // Mat frame, blob;
   std::vector<int> indices;
 
   Rcpp::IntegerVector x1vec(indices.size());
@@ -98,14 +87,8 @@ Rcpp::DataFrame
   Rcpp::IntegerVector y4vec(indices.size());
 
   // if window is drawn, push super key to exit
-  while (waitKey(1) < 0)
-  {
-    cap >> frame;
-    if (frame.empty())
-    {
-      waitKey();
-      break;
-    }
+  // while (waitKey(1) < 0)
+  // {
 
     blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
                   Scalar(123.68, 116.78, 103.94), true, false);
@@ -163,16 +146,11 @@ Rcpp::DataFrame
               FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
     }
 
-    if (draw)
-      imshow(kWinName, frame);
-    else
-      break;
-  }
+  //     break;
+  // }
 
-  // XPtrMat out = cvmat_xptr(box);
-  // out.attr("indices") = ;
-
-  return Rcpp::DataFrame::create(
+  XPtrMat out = cvmat_xptr(frame);
+  out.attr("indices") =  Rcpp::DataFrame::create(
     Rcpp::_["x1"] = x1vec,
     Rcpp::_["y1"] = y1vec,
     Rcpp::_["x2"] = x2vec,
@@ -182,4 +160,6 @@ Rcpp::DataFrame
     Rcpp::_["x4"] = x4vec,
     Rcpp::_["y4"] = y4vec
   );
+
+  return out;
 }

From d2b847c0d7fe090abd911114bd432c7f0b738c6a Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Sun, 5 Apr 2020 04:19:44 +0200
Subject: [PATCH 03/10] cleanup

---
 src/text.cpp | 96 ++++++++++++++++++++++++----------------------------
 1 file changed, 45 insertions(+), 51 deletions(-)

diff --git a/src/text.cpp b/src/text.cpp
index 0c26f4e..d2b151e 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -59,36 +59,24 @@ void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
 XPtrMat
   text_detection(XPtrMat ptr, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
-{
-  if (model.empty())
-    Rcpp::stop("No model defined");
-
-  Mat frame = get_mat(ptr);
+  {
+    if (model.empty())
+      Rcpp::stop("No model defined");
 
-  // Load network.
-  Net net = readNet(model);
+    Mat frame = get_mat(ptr);
 
-  std::vector<Mat> outs;
-  std::vector<String> outNames(2);
-  outNames[0] = "feature_fusion/Conv_7/Sigmoid";
-  outNames[1] = "feature_fusion/concat_3";
+    // Load network.
+    Net net = readNet(model);
 
-  Mat blob;
-  // Mat frame, blob;
-  std::vector<int> indices;
+    std::vector<Mat> outs;
+    std::vector<String> outNames(2);
+    outNames[0] = "feature_fusion/Conv_7/Sigmoid";
+    outNames[1] = "feature_fusion/concat_3";
 
-  Rcpp::IntegerVector x1vec(indices.size());
-  Rcpp::IntegerVector y1vec(indices.size());
-  Rcpp::IntegerVector x2vec(indices.size());
-  Rcpp::IntegerVector y2vec(indices.size());
-  Rcpp::IntegerVector x3vec(indices.size());
-  Rcpp::IntegerVector y3vec(indices.size());
-  Rcpp::IntegerVector x4vec(indices.size());
-  Rcpp::IntegerVector y4vec(indices.size());
+    Mat blob;
+    // Mat frame, blob;
+    std::vector<int> indices;
 
-  // if window is drawn, push super key to exit
-  // while (waitKey(1) < 0)
-  // {
 
     blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
                   Scalar(123.68, 116.78, 103.94), true, false);
@@ -106,6 +94,15 @@ XPtrMat
     // Apply non-maximum suppression procedure.
     NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
 
+    Rcpp::IntegerVector x1vec(indices.size());
+    Rcpp::IntegerVector y1vec(indices.size());
+    Rcpp::IntegerVector x2vec(indices.size());
+    Rcpp::IntegerVector y2vec(indices.size());
+    Rcpp::IntegerVector x3vec(indices.size());
+    Rcpp::IntegerVector y3vec(indices.size());
+    Rcpp::IntegerVector x4vec(indices.size());
+    Rcpp::IntegerVector y4vec(indices.size());
+
 
     // Render detections.
     Point2f ratio((float)frame.cols / inpWidth, (float)frame.rows / inpHeight);
@@ -121,15 +118,14 @@ XPtrMat
         vertices[j].y *= ratio.y;
       }
 
-      // not the fastes way
-      x1vec.push_back(vertices[0].x);
-      y1vec.push_back(vertices[0].y);
-      x2vec.push_back(vertices[1].x);
-      y2vec.push_back(vertices[1].y);
-      x3vec.push_back(vertices[2].x);
-      y3vec.push_back(vertices[2].y);
-      x4vec.push_back(vertices[3].x);
-      y4vec.push_back(vertices[3].y);
+      x1vec.at(i) = vertices[0].x;
+      y1vec.at(i) = vertices[0].y;
+      x2vec.at(i) = vertices[1].x;
+      y2vec.at(i) = vertices[1].y;
+      x3vec.at(i) = vertices[2].x;
+      y3vec.at(i) = vertices[2].y;
+      x4vec.at(i) = vertices[3].x;
+      y4vec.at(i) = vertices[3].y;
 
       if (draw)
         for (int j = 0; j < 4; ++j)
@@ -146,20 +142,18 @@ XPtrMat
               FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
     }
 
-  //     break;
-  // }
-
-  XPtrMat out = cvmat_xptr(frame);
-  out.attr("indices") =  Rcpp::DataFrame::create(
-    Rcpp::_["x1"] = x1vec,
-    Rcpp::_["y1"] = y1vec,
-    Rcpp::_["x2"] = x2vec,
-    Rcpp::_["y2"] = y2vec,
-    Rcpp::_["x3"] = x3vec,
-    Rcpp::_["y3"] = y3vec,
-    Rcpp::_["x4"] = x4vec,
-    Rcpp::_["y4"] = y4vec
-  );
-
-  return out;
-}
+
+    XPtrMat out = cvmat_xptr(frame);
+    out.attr("indices") =  Rcpp::DataFrame::create(
+      Rcpp::_["x1"] = x1vec,
+      Rcpp::_["y1"] = y1vec,
+      Rcpp::_["x2"] = x2vec,
+      Rcpp::_["y2"] = y2vec,
+      Rcpp::_["x3"] = x3vec,
+      Rcpp::_["y3"] = y3vec,
+      Rcpp::_["x4"] = x4vec,
+      Rcpp::_["y4"] = y4vec
+    );
+
+    return out;
+  }

From 7040d7461cf389b1918b8a9684c4f2bb1042d148 Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Sun, 5 Apr 2020 13:49:03 +0200
Subject: [PATCH 04/10] cleanups

---
 R/RcppExports.R     |  4 ++--
 R/text.R            |  2 +-
 src/RcppExports.cpp |  8 ++++----
 src/text.cpp        | 10 +++++-----
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/R/RcppExports.R b/R/RcppExports.R
index 7897015..f14a751 100644
--- a/R/RcppExports.R
+++ b/R/RcppExports.R
@@ -105,7 +105,7 @@ cvmat_markers <- function(ptr) {
     .Call('_opencv_cvmat_markers', PACKAGE = 'opencv', ptr)
 }
 
-text_detection <- function(ptr, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw) {
-    .Call('_opencv_text_detection', PACKAGE = 'opencv', ptr, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw)
+text_detection <- function(input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw) {
+    .Call('_opencv_text_detection', PACKAGE = 'opencv', input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw)
 }
 
diff --git a/R/text.R b/R/text.R
index 3a18a74..869018e 100644
--- a/R/text.R
+++ b/R/text.R
@@ -15,7 +15,7 @@ ocv_text <- function(image, thrs = 0.5, nms = 20, width = 320, height = 320,
 
   model <- path.expand(model)
 
-  text_detection(ptr = image,
+  text_detection(input = image,
                  confThreshold = thrs,
                  nmsThreshold = nms,
                  inpWidth = width,
diff --git a/src/RcppExports.cpp b/src/RcppExports.cpp
index 9829f59..e606488 100644
--- a/src/RcppExports.cpp
+++ b/src/RcppExports.cpp
@@ -297,19 +297,19 @@ BEGIN_RCPP
 END_RCPP
 }
 // text_detection
-XPtrMat text_detection(XPtrMat ptr, float confThreshold, float nmsThreshold, int inpWidth, int inpHeight, std::string model, bool draw);
-RcppExport SEXP _opencv_text_detection(SEXP ptrSEXP, SEXP confThresholdSEXP, SEXP nmsThresholdSEXP, SEXP inpWidthSEXP, SEXP inpHeightSEXP, SEXP modelSEXP, SEXP drawSEXP) {
+XPtrMat text_detection(XPtrMat input, float confThreshold, float nmsThreshold, int inpWidth, int inpHeight, std::string model, bool draw);
+RcppExport SEXP _opencv_text_detection(SEXP inputSEXP, SEXP confThresholdSEXP, SEXP nmsThresholdSEXP, SEXP inpWidthSEXP, SEXP inpHeightSEXP, SEXP modelSEXP, SEXP drawSEXP) {
 BEGIN_RCPP
     Rcpp::RObject rcpp_result_gen;
     Rcpp::RNGScope rcpp_rngScope_gen;
-    Rcpp::traits::input_parameter< XPtrMat >::type ptr(ptrSEXP);
+    Rcpp::traits::input_parameter< XPtrMat >::type input(inputSEXP);
     Rcpp::traits::input_parameter< float >::type confThreshold(confThresholdSEXP);
     Rcpp::traits::input_parameter< float >::type nmsThreshold(nmsThresholdSEXP);
     Rcpp::traits::input_parameter< int >::type inpWidth(inpWidthSEXP);
     Rcpp::traits::input_parameter< int >::type inpHeight(inpHeightSEXP);
     Rcpp::traits::input_parameter< std::string >::type model(modelSEXP);
     Rcpp::traits::input_parameter< bool >::type draw(drawSEXP);
-    rcpp_result_gen = Rcpp::wrap(text_detection(ptr, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw));
+    rcpp_result_gen = Rcpp::wrap(text_detection(input, confThreshold, nmsThreshold, inpWidth, inpHeight, model, draw));
     return rcpp_result_gen;
 END_RCPP
 }
diff --git a/src/text.cpp b/src/text.cpp
index d2b151e..05f0540 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -1,6 +1,5 @@
 #include "util.hpp"
 #include <opencv2/imgproc.hpp>
-#include <opencv2/highgui.hpp>
 #include <opencv2/dnn.hpp>
 
 using namespace cv;
@@ -57,13 +56,14 @@ void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
 
 // [[Rcpp::export]]
 XPtrMat
-  text_detection(XPtrMat ptr, float confThreshold,float nmsThreshold,
+  text_detection(XPtrMat input, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
   {
     if (model.empty())
       Rcpp::stop("No model defined");
 
-    Mat frame = get_mat(ptr);
+    Mat inp = get_mat(input);
+
 
     // Load network.
     Net net = readNet(model);
@@ -73,10 +73,10 @@ XPtrMat
     outNames[0] = "feature_fusion/Conv_7/Sigmoid";
     outNames[1] = "feature_fusion/concat_3";
 
-    Mat blob;
-    // Mat frame, blob;
+    Mat frame, blob;
     std::vector<int> indices;
 
+    frame = inp.clone();
 
     blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
                   Scalar(123.68, 116.78, 103.94), true, false);

From 3cf07e14aac84b1a1dcc7e6842c5157334994de8 Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Sun, 5 Apr 2020 14:26:14 +0200
Subject: [PATCH 05/10] cleanup

---
 R/text.R        | 34 +++++++++++++++++++++++++------
 man/ocv_text.Rd | 53 +++++++++++++++++++++++++++++++++++++++++++++++++
 man/opencv.Rd   |  5 +----
 3 files changed, 82 insertions(+), 10 deletions(-)
 create mode 100644 man/ocv_text.Rd

diff --git a/R/text.R b/R/text.R
index 869018e..902883b 100644
--- a/R/text.R
+++ b/R/text.R
@@ -1,12 +1,34 @@
-#' @export
-#' @rdname opencv
-#' @param image Path to input image or video file. Skip this argument to capture frames from a camera.
-#' @param width Preprocess input image by resizing to a specific width. It should be multiple by 32.
-#' @param height Preprocess input image by resizing to a specific height. It should be multiple by 32.
+#' ocv_text text detection in images using EAST
+#'
+#' @description OpenCV sample using EAST (Efficient and Accurate Scene Text
+#'   Detector) to detect text in images. Requires the EAST pb-model not included
+#'   in the package.
+#'
+#' @param image Path to input image or video file. Skip this argument to capture
+#'   frames from a camera.
+#' @param width Preprocess input image by resizing to a specific width. It
+#'   should be multiple by 32.
+#' @param height Preprocess input image by resizing to a specific height. It
+#'   should be multiple by 32.
 #' @param thrs Confidence threshold.
 #' @param nms Non-maximum suppression threshold.
-#'
 #' @param model Path to a binary .pb file contains trained network.
+#' @param draw Draws visual output to the image.
+#'
+#' @examples
+#' \dontrun{
+#' url <- paste0('https://upload.wikimedia.org/wikipedia/commons/6/6f/',
+#'               'Keep-calm-and-carry-on-scan.jpg')
+#' fl <- ocv_read(url)
+#'
+#' ocv_text(image = fl, thrs =  0.7,  nms = 0.3,
+#'          model = "frozen_east_text_detection.pb")
+#' }
+#'
+#' @references
+#' https://github.com/opencv/opencv/blob/master/samples/dnn/text_detection.cpp
+#'
+#' @export
 ocv_text <- function(image, thrs = 0.5, nms = 20, width = 320, height = 320,
                      model, draw = FALSE){
 
diff --git a/man/ocv_text.Rd b/man/ocv_text.Rd
new file mode 100644
index 0000000..c9b193c
--- /dev/null
+++ b/man/ocv_text.Rd
@@ -0,0 +1,53 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/text.R
+\name{ocv_text}
+\alias{ocv_text}
+\title{ocv_text text detection in images using EAST}
+\usage{
+ocv_text(
+  image,
+  thrs = 0.5,
+  nms = 20,
+  width = 320,
+  height = 320,
+  model,
+  draw = FALSE
+)
+}
+\arguments{
+\item{image}{Path to input image or video file. Skip this argument to capture
+frames from a camera.}
+
+\item{thrs}{Confidence threshold.}
+
+\item{nms}{Non-maximum suppression threshold.}
+
+\item{width}{Preprocess input image by resizing to a specific width. It
+should be multiple by 32.}
+
+\item{height}{Preprocess input image by resizing to a specific height. It
+should be multiple by 32.}
+
+\item{model}{Path to a binary .pb file contains trained network.}
+
+\item{draw}{Draws visual output to the image.}
+}
+\description{
+OpenCV sample using EAST (Efficient and Accurate Scene Text
+Detector) to detect text in images. Requires the EAST pb-model not included
+in the package.
+}
+\examples{
+\dontrun{
+url <- paste0('https://upload.wikimedia.org/wikipedia/commons/6/6f/',
+              'Keep-calm-and-carry-on-scan.jpg')
+fl <- ocv_read(url)
+
+ocv_text(image = fl, thrs =  0.7,  nms = 0.3,
+         model = "frozen_east_text_detection.pb")
+}
+
+}
+\references{
+https://github.com/opencv/opencv/blob/master/samples/dnn/text_detection.cpp
+}
diff --git a/man/opencv.Rd b/man/opencv.Rd
index 3daaf0b..f88d0f5 100644
--- a/man/opencv.Rd
+++ b/man/opencv.Rd
@@ -1,5 +1,5 @@
 % Generated by roxygen2: do not edit by hand
-% Please edit documentation in R/face.R, R/opencv.R, R/text.R
+% Please edit documentation in R/face.R, R/opencv.R
 \name{ocv_face}
 \alias{ocv_face}
 \alias{ocv_facemask}
@@ -21,7 +21,6 @@
 \alias{ocv_copyto}
 \alias{ocv_display}
 \alias{ocv_video}
-\alias{ocv_text}
 \title{OpenCV Computer Vision}
 \usage{
 ocv_face(image)
@@ -63,8 +62,6 @@ ocv_copyto(image, target, mask)
 ocv_display(image)
 
 ocv_video(filter)
-
-ocv_text(image, thrs = 0.5, nms = 20, width = 320, height = 320, model)
 }
 \arguments{
 \item{image}{a ocv image object}

From 422f7c0802900a5dd4825506be7fce356c08b6ff Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Mon, 6 Apr 2020 20:10:57 +0200
Subject: [PATCH 06/10] dnn requires 3.4.3

---
 src/text.cpp | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/src/text.cpp b/src/text.cpp
index 05f0540..e285afb 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -9,6 +9,9 @@ void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
             std::vector<RotatedRect>& detections,
             std::vector<float>& confidences)
 {
+#if CV_VERSION_MAJOR < 3 and CV_VERSION_MINOR < 4 and  CV_VERSION_REVISION < 3
+  throw std::runtime_error("createBackgroundSubtractorMOG2 requires OpenCV 3 or newer");
+#else
   detections.clear();
   CV_Assert(scores.dims == 4); CV_Assert(geometry.dims == 4);
   CV_Assert(scores.size[0] == 1); CV_Assert(geometry.size[0] == 1);
@@ -51,6 +54,7 @@ void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
       confidences.push_back(score);
     }
   }
+#endif
 }
 
 
@@ -59,6 +63,9 @@ XPtrMat
   text_detection(XPtrMat input, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
   {
+#if CV_VERSION_MAJOR < 3 and CV_VERSION_MINOR < 4 and  CV_VERSION_REVISION < 3
+    throw std::runtime_error("createBackgroundSubtractorMOG2 requires OpenCV 3 or newer");
+#else
     if (model.empty())
       Rcpp::stop("No model defined");
 
@@ -156,4 +163,5 @@ XPtrMat
     );
 
     return out;
+#endif
   }

From fb28f9e5fc24e6a26a818a36c3a01a3b48856e62 Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Mon, 6 Apr 2020 20:48:41 +0200
Subject: [PATCH 07/10] dnn requires 3.4.3

---
 src/text.cpp | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/src/text.cpp b/src/text.cpp
index e285afb..7cfd654 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -1,16 +1,19 @@
 #include "util.hpp"
 #include <opencv2/imgproc.hpp>
+
+#if CV_VERSION_MAJOR >= 3 && CV_VERSION_MINOR >= 4 &&  CV_VERSION_REVISION >= 3
 #include <opencv2/dnn.hpp>
+#endif
 
 using namespace cv;
-using namespace cv::dnn;
+// using namespace cv::dnn;
 
 void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
             std::vector<RotatedRect>& detections,
             std::vector<float>& confidences)
 {
-#if CV_VERSION_MAJOR < 3 and CV_VERSION_MINOR < 4 and  CV_VERSION_REVISION < 3
-  throw std::runtime_error("createBackgroundSubtractorMOG2 requires OpenCV 3 or newer");
+#if CV_VERSION_MAJOR < 3 && CV_VERSION_MINOR < 4 &&  CV_VERSION_REVISION < 3
+  throw std::runtime_error("ocv_text requires OpenCV 3.4.3 or newer");
 #else
   detections.clear();
   CV_Assert(scores.dims == 4); CV_Assert(geometry.dims == 4);
@@ -63,8 +66,8 @@ XPtrMat
   text_detection(XPtrMat input, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
   {
-#if CV_VERSION_MAJOR < 3 and CV_VERSION_MINOR < 4 and  CV_VERSION_REVISION < 3
-    throw std::runtime_error("createBackgroundSubtractorMOG2 requires OpenCV 3 or newer");
+#if CV_VERSION_MAJOR < 3 && CV_VERSION_MINOR < 4 &&  CV_VERSION_REVISION < 3
+    throw std::runtime_error("ocv_text requires OpenCV 3.4.3 or newer");
 #else
     if (model.empty())
       Rcpp::stop("No model defined");
@@ -73,7 +76,7 @@ XPtrMat
 
 
     // Load network.
-    Net net = readNet(model);
+    cv::dnn::Net net = cv::dnn::readNet(model);
 
     std::vector<Mat> outs;
     std::vector<String> outNames(2);
@@ -85,8 +88,8 @@ XPtrMat
 
     frame = inp.clone();
 
-    blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
-                  Scalar(123.68, 116.78, 103.94), true, false);
+    cv::dnn::blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
+                           Scalar(123.68, 116.78, 103.94), true, false);
     net.setInput(blob);
     net.forward(outs, outNames);
 
@@ -99,7 +102,8 @@ XPtrMat
     decode(scores, geometry, confThreshold, boxes, confidences);
 
     // Apply non-maximum suppression procedure.
-    NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
+
+    cv::dnn::NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
 
     Rcpp::IntegerVector x1vec(indices.size());
     Rcpp::IntegerVector y1vec(indices.size());

From 0c986eee126d2003bf4e35f5cbe92178cf288421 Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Fri, 10 Apr 2020 13:59:30 +0200
Subject: [PATCH 08/10] dnn requires 3.4.3 or newer with DNN

---
 src/text.cpp | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/src/text.cpp b/src/text.cpp
index 7cfd654..cd519c0 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -1,19 +1,18 @@
 #include "util.hpp"
 #include <opencv2/imgproc.hpp>
 
-#if CV_VERSION_MAJOR >= 3 && CV_VERSION_MINOR >= 4 &&  CV_VERSION_REVISION >= 3
-#include <opencv2/dnn.hpp>
+#ifdef HAVE_OPENCV_DNN
+#include "opencv2/dnn.hpp"
 #endif
 
 using namespace cv;
-// using namespace cv::dnn;
 
 void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
             std::vector<RotatedRect>& detections,
             std::vector<float>& confidences)
 {
-#if CV_VERSION_MAJOR < 3 && CV_VERSION_MINOR < 4 &&  CV_VERSION_REVISION < 3
-  throw std::runtime_error("ocv_text requires OpenCV 3.4.3 or newer");
+#ifndef HAVE_OPENCV_DNN
+  throw std::runtime_error("ocv_text req. OpenCV 3.4.3 or newer with DNN");
 #else
   detections.clear();
   CV_Assert(scores.dims == 4); CV_Assert(geometry.dims == 4);
@@ -66,8 +65,8 @@ XPtrMat
   text_detection(XPtrMat input, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
   {
-#if CV_VERSION_MAJOR < 3 && CV_VERSION_MINOR < 4 &&  CV_VERSION_REVISION < 3
-    throw std::runtime_error("ocv_text requires OpenCV 3.4.3 or newer");
+#ifndef HAVE_OPENCV_DNN
+    throw std::runtime_error("ocv_text req. OpenCV 3.4.3 or newer with DNN");
 #else
     if (model.empty())
       Rcpp::stop("No model defined");

From bf708c03d3c6be472a08035f900a27b08652a865 Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Fri, 10 Apr 2020 13:59:30 +0200
Subject: [PATCH 09/10] dnn requires 3.4.3 or newer with DNN

---
 src/text.cpp | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)

diff --git a/src/text.cpp b/src/text.cpp
index 7cfd654..33bb6f8 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -1,19 +1,22 @@
 #include "util.hpp"
 #include <opencv2/imgproc.hpp>
 
-#if CV_VERSION_MAJOR >= 3 && CV_VERSION_MINOR >= 4 &&  CV_VERSION_REVISION >= 3
-#include <opencv2/dnn.hpp>
+#define OPENCV_VERSION (CV_VERSION_MAJOR * 10000 \
++ CV_VERSION_MINOR * 100                         \
++ CV_VERSION_REVISION)
+
+#ifdef HAVE_OPENCV_DNN
+#include "opencv2/dnn.hpp"
 #endif
 
 using namespace cv;
-// using namespace cv::dnn;
 
 void decode(const Mat& scores, const Mat& geometry, float scoreThresh,
             std::vector<RotatedRect>& detections,
             std::vector<float>& confidences)
 {
-#if CV_VERSION_MAJOR < 3 && CV_VERSION_MINOR < 4 &&  CV_VERSION_REVISION < 3
-  throw std::runtime_error("ocv_text requires OpenCV 3.4.3 or newer");
+#if !defined(HAVE_OPENCV_DNN) || OPENCV_VERSION < 30403
+  throw std::runtime_error("ocv_text req. OpenCV 3.4.3 or newer with DNN");
 #else
   detections.clear();
   CV_Assert(scores.dims == 4); CV_Assert(geometry.dims == 4);
@@ -66,8 +69,8 @@ XPtrMat
   text_detection(XPtrMat input, float confThreshold,float nmsThreshold,
                  int inpWidth, int inpHeight, std::string model, bool draw)
   {
-#if CV_VERSION_MAJOR < 3 && CV_VERSION_MINOR < 4 &&  CV_VERSION_REVISION < 3
-    throw std::runtime_error("ocv_text requires OpenCV 3.4.3 or newer");
+#if !defined(HAVE_OPENCV_DNN) || OPENCV_VERSION < 30403
+    throw std::runtime_error("ocv_text req. OpenCV 3.4.3 or newer with DNN");
 #else
     if (model.empty())
       Rcpp::stop("No model defined");

From 653ee42e65ab0d5a3d61165f2558d9eb9cb05ad6 Mon Sep 17 00:00:00 2001
From: Jan Marvin Garbuszus <jan.garbuszus@ruhr-uni-bochum.de>
Date: Fri, 10 Apr 2020 15:09:01 +0200
Subject: [PATCH 10/10] clean up

---
 R/text.R        |  5 ++---
 man/ocv_text.Rd |  5 ++---
 src/text.cpp    | 23 ++++++++++-------------
 3 files changed, 14 insertions(+), 19 deletions(-)

diff --git a/R/text.R b/R/text.R
index 902883b..b12a3b2 100644
--- a/R/text.R
+++ b/R/text.R
@@ -4,13 +4,12 @@
 #'   Detector) to detect text in images. Requires the EAST pb-model not included
 #'   in the package.
 #'
-#' @param image Path to input image or video file. Skip this argument to capture
-#'   frames from a camera.
+#' @param image opencv image to be processed.
 #' @param width Preprocess input image by resizing to a specific width. It
 #'   should be multiple by 32.
 #' @param height Preprocess input image by resizing to a specific height. It
 #'   should be multiple by 32.
-#' @param thrs Confidence threshold.
+#' @param thrs Confidence threshold between 0 and 1.
 #' @param nms Non-maximum suppression threshold.
 #' @param model Path to a binary .pb file contains trained network.
 #' @param draw Draws visual output to the image.
diff --git a/man/ocv_text.Rd b/man/ocv_text.Rd
index c9b193c..475bf78 100644
--- a/man/ocv_text.Rd
+++ b/man/ocv_text.Rd
@@ -15,10 +15,9 @@ ocv_text(
 )
 }
 \arguments{
-\item{image}{Path to input image or video file. Skip this argument to capture
-frames from a camera.}
+\item{image}{opencv image to be processed.}
 
-\item{thrs}{Confidence threshold.}
+\item{thrs}{Confidence threshold between 0 and 1.}
 
 \item{nms}{Non-maximum suppression threshold.}
 
diff --git a/src/text.cpp b/src/text.cpp
index 33bb6f8..a25c7e0 100644
--- a/src/text.cpp
+++ b/src/text.cpp
@@ -72,24 +72,21 @@ XPtrMat
 #if !defined(HAVE_OPENCV_DNN) || OPENCV_VERSION < 30403
     throw std::runtime_error("ocv_text req. OpenCV 3.4.3 or newer with DNN");
 #else
-    if (model.empty())
-      Rcpp::stop("No model defined");
 
-    Mat inp = get_mat(input);
 
+    Mat frame, blob;
+    Mat inp = get_mat(input);
+    frame = inp.clone();
 
-    // Load network.
-    cv::dnn::Net net = cv::dnn::readNet(model);
+    std::vector<int> indices;
 
     std::vector<Mat> outs;
     std::vector<String> outNames(2);
     outNames[0] = "feature_fusion/Conv_7/Sigmoid";
     outNames[1] = "feature_fusion/concat_3";
 
-    Mat frame, blob;
-    std::vector<int> indices;
-
-    frame = inp.clone();
+    // Load network.
+    cv::dnn::Net net = cv::dnn::readNet(model);
 
     cv::dnn::blobFromImage(frame, blob, 1.0, Size(inpWidth, inpHeight),
                            Scalar(123.68, 116.78, 103.94), true, false);
@@ -105,9 +102,9 @@ XPtrMat
     decode(scores, geometry, confThreshold, boxes, confidences);
 
     // Apply non-maximum suppression procedure.
-
     cv::dnn::NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
 
+    // initiate output vetors
     Rcpp::IntegerVector x1vec(indices.size());
     Rcpp::IntegerVector y1vec(indices.size());
     Rcpp::IntegerVector x2vec(indices.size());
@@ -117,7 +114,6 @@ XPtrMat
     Rcpp::IntegerVector x4vec(indices.size());
     Rcpp::IntegerVector y4vec(indices.size());
 
-
     // Render detections.
     Point2f ratio((float)frame.cols / inpWidth, (float)frame.rows / inpHeight);
     for (size_t i = 0; i < indices.size(); ++i)
@@ -141,13 +137,14 @@ XPtrMat
       x4vec.at(i) = vertices[3].x;
       y4vec.at(i) = vertices[3].y;
 
+      // draws boxes to image
       if (draw)
         for (int j = 0; j < 4; ++j)
           line(frame, vertices[j], vertices[(j + 1) % 4], Scalar(0, 255, 0), 1);
     }
 
     if (draw) {
-      // Put efficiency information.
+      // Draws efficiency information to the image.
       std::vector<double> layersTimes;
       double freq = getTickFrequency() / 1000;
       double t = net.getPerfProfile(layersTimes) / freq;
@@ -156,7 +153,7 @@ XPtrMat
               FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
     }
 
-
+    // prepare output
     XPtrMat out = cvmat_xptr(frame);
     out.attr("indices") =  Rcpp::DataFrame::create(
       Rcpp::_["x1"] = x1vec,