Merge pull request #9820 from sovrasov:text_detector_dnn

modules/dnn/misc/caffe/caffe.pb.h

@@ -1537,6 +1537,13 @@ class PriorBoxParameter : public ::google::protobuf::Message /* @@protoc_inserti
   float offset() const;
   void set_offset(float value);
 
+  // optional bool additional_y_offset = 14 [default = false];
+  bool has_additional_y_offset() const;
+  void clear_additional_y_offset();
+  static const int kAdditionalYOffsetFieldNumber = 14;
+  bool additional_y_offset() const;
+  void set_additional_y_offset(bool value);
+
   // @@protoc_insertion_point(class_scope:caffe.PriorBoxParameter)
  private:
   inline void set_has_min_size();
@@ -1561,6 +1568,8 @@ class PriorBoxParameter : public ::google::protobuf::Message /* @@protoc_inserti
   inline void clear_has_step_w();
   inline void set_has_offset();
   inline void clear_has_offset();
+  inline void set_has_additional_y_offset();
+  inline void clear_has_additional_y_offset();
 
   ::google::protobuf::internal::InternalMetadataWithArena _internal_metadata_;
   ::google::protobuf::internal::HasBits<1> _has_bits_;
@@ -1575,6 +1584,7 @@ class PriorBoxParameter : public ::google::protobuf::Message /* @@protoc_inserti
   float step_;
   float step_h_;
   float step_w_;
+  bool additional_y_offset_;
   bool flip_;
   bool clip_;
   float offset_;
@@ -13635,6 +13645,30 @@ inline void PriorBoxParameter::set_offset(float value) {
   // @@protoc_insertion_point(field_set:caffe.PriorBoxParameter.offset)
 }
 
+// optional bool additional_y_offset = 14 [default = false];
+inline bool PriorBoxParameter::has_additional_y_offset() const {
+  return (_has_bits_[0] & 0x00002000u) != 0;
+}
+inline void PriorBoxParameter::set_has_additional_y_offset() {
+  _has_bits_[0] |= 0x00002000u;
+}
+inline void PriorBoxParameter::clear_has_additional_y_offset() {
+  _has_bits_[0] &= ~0x00002000u;
+}
+inline void PriorBoxParameter::clear_additional_y_offset() {
+  additional_y_offset_ = false;
+  clear_has_additional_y_offset();
+}
+inline bool PriorBoxParameter::additional_y_offset() const {
+  // @@protoc_insertion_point(field_get:caffe.PriorBoxParameter.additional_y_offset)
+  return additional_y_offset_;
+}
+inline void PriorBoxParameter::set_additional_y_offset(bool value) {
+  set_has_additional_y_offset();
+  additional_y_offset_ = value;
+  // @@protoc_insertion_point(field_set:caffe.PriorBoxParameter.additional_y_offset)
+}
+
 inline const PriorBoxParameter* PriorBoxParameter::internal_default_instance() {
   return &PriorBoxParameter_default_instance_.get();
 }

modules/dnn/src/caffe/caffe.proto

@@ -145,6 +145,8 @@ message PriorBoxParameter {
   optional float step_w = 12;
   // Offset to the top left corner of each cell.
   optional float offset = 13 [default = 0.5];
+  // If true, two additional boxes for each center will be generated. Their centers will be shifted by y coordinate.
+  optional bool additional_y_offset = 14 [default = false];
 }
 
 // Message that store parameters used by DetectionOutputLayer

modules/dnn/src/layers/prior_box_layer.cpp

@@ -216,6 +216,14 @@ public:
           _stepY = 0;
           _stepX = 0;
         }
+        if(params.has("additional_y_offset"))
+        {
+          _additional_y_offset = getParameter<bool>(params, "additional_y_offset");
+          if(_additional_y_offset)
+            _numPriors *= 2;
+        }
+        else
+          _additional_y_offset = false;
     }
 
     bool getMemoryShapes(const std::vector<MatShape> &inputs,
@@ -246,10 +254,11 @@ public:
         CV_TRACE_FUNCTION();
         CV_TRACE_ARG_VALUE(name, "name", name.c_str());
 
+        size_t real_numPriors = _additional_y_offset ? _numPriors / 2 : _numPriors;
         if (_scales.empty())
-            _scales.resize(_numPriors, 1.0f);
+            _scales.resize(real_numPriors, 1.0f);
         else
-            CV_Assert(_scales.size() == _numPriors);
+            CV_Assert(_scales.size() == real_numPriors);
 
         int _layerWidth = inputs[0]->size[3];
         int _layerHeight = inputs[0]->size[2];
@@ -289,6 +298,19 @@ public:
                 // ymax
                 outputPtr[idx++] = (center_y + _boxHeight / 2.) / _imageHeight;
 
+                if(_additional_y_offset)
+                {
+                  float center_y_offset_1 = (h + 1.0) * stepY;
+                  // xmin
+                  outputPtr[idx++] = (center_x - _boxWidth / 2.) / _imageWidth;
+                  // ymin
+                  outputPtr[idx++] = (center_y_offset_1 - _boxHeight / 2.) / _imageHeight;
+                  // xmax
+                  outputPtr[idx++] = (center_x + _boxWidth / 2.) / _imageWidth;
+                  // ymax
+                  outputPtr[idx++] = (center_y_offset_1 + _boxHeight / 2.) / _imageHeight;
+                }
+
                 if (_maxSize > 0)
                 {
                     // second prior: aspect_ratio = 1, size = sqrt(min_size * max_size)
@@ -301,6 +323,19 @@ public:
                     outputPtr[idx++] = (center_x + _boxWidth / 2.) / _imageWidth;
                     // ymax
                     outputPtr[idx++] = (center_y + _boxHeight / 2.) / _imageHeight;
+
+                    if(_additional_y_offset)
+                    {
+                      float center_y_offset_1 = (h + 1.0) * stepY;
+                      // xmin
+                      outputPtr[idx++] = (center_x - _boxWidth / 2.) / _imageWidth;
+                      // ymin
+                      outputPtr[idx++] = (center_y_offset_1 - _boxHeight / 2.) / _imageHeight;
+                      // xmax
+                      outputPtr[idx++] = (center_x + _boxWidth / 2.) / _imageWidth;
+                      // ymax
+                      outputPtr[idx++] = (center_y_offset_1 + _boxHeight / 2.) / _imageHeight;
+                    }
                 }
 
                 // rest of priors
@@ -319,6 +354,18 @@ public:
                     outputPtr[idx++] = (center_x + _boxWidth / 2.) / _imageWidth;
                     // ymax
                     outputPtr[idx++] = (center_y + _boxHeight / 2.) / _imageHeight;
+                    if(_additional_y_offset)
+                    {
+                      float center_y_offset_1 = (h + 1.0) * stepY;
+                      // xmin
+                      outputPtr[idx++] = (center_x - _boxWidth / 2.) / _imageWidth;
+                      // ymin
+                      outputPtr[idx++] = (center_y_offset_1 - _boxHeight / 2.) / _imageHeight;
+                      // xmax
+                      outputPtr[idx++] = (center_x + _boxWidth / 2.) / _imageWidth;
+                      // ymax
+                      outputPtr[idx++] = (center_y_offset_1 + _boxHeight / 2.) / _imageHeight;
+                    }
                 }
             }
         }
@@ -385,6 +432,7 @@ public:
 
     bool _flip;
     bool _clip;
+    bool _additional_y_offset;
 
     size_t _numPriors;