opencv: facedetect: add right/left ear detection

2 files changed, 78 insertions, 4 deletions

diff --git a/ext/opencv/gstfacedetect.cpp b/ext/opencv/gstfacedetect.cpp
index c7cb4d538..1b0d2d474 100644
--- a/ext/opencv/gstfacedetect.cpp
+++ b/ext/opencv/gstfacedetect.cpp
@@ -94,6 +94,8 @@ GST_DEBUG_CATEGORY_STATIC (gst_face_detect_debug);
 #define DEFAULT_NOSE_PROFILE "haarcascade_mcs_nose.xml"
 #define DEFAULT_MOUTH_PROFILE "haarcascade_mcs_mouth.xml"
 #define DEFAULT_EYES_PROFILE "haarcascade_mcs_eyepair_small.xml"
+#define DEFAULT_EAR_L_PROFILE "haarcascade_mcs_leftear.xml"
+#define DEFAULT_EAR_R_PROFILE "haarcascade_mcs_rightear.xml"
 #define DEFAULT_SCALE_FACTOR 1.25
 #if (CV_MAJOR_VERSION >= 4)
 #define DEFAULT_FLAGS CASCADE_DO_CANNY_PRUNING
@@ -254,6 +256,8 @@ gst_face_detect_finalize (GObject * obj)
   g_free (filter->nose_profile);
   g_free (filter->mouth_profile);
   g_free (filter->eyes_profile);
+  g_free (filter->ear_l_profile);
+  g_free (filter->ear_r_profile);
 
   if (filter->cvFaceDetect)
     delete (filter->cvFaceDetect);
@@ -263,6 +267,10 @@ gst_face_detect_finalize (GObject * obj)
     delete (filter->cvMouthDetect);
   if (filter->cvEyesDetect)
     delete (filter->cvEyesDetect);
+  if (filter->cvEarLDetect)
+    delete (filter->cvEarLDetect);
+  if (filter->cvEarRDetect)
+    delete (filter->cvEarRDetect);
 
   G_OBJECT_CLASS (gst_face_detect_parent_class)->finalize (obj);
 }
@@ -374,6 +382,8 @@ gst_face_detect_init (GstFaceDetect * filter)
   filter->nose_profile = g_strconcat (haar_path, DEFAULT_NOSE_PROFILE, NULL);
   filter->mouth_profile = g_strconcat (haar_path, DEFAULT_MOUTH_PROFILE, NULL);
   filter->eyes_profile = g_strconcat (haar_path, DEFAULT_EYES_PROFILE, NULL);
+  filter->ear_r_profile = g_strconcat (haar_path, DEFAULT_EAR_R_PROFILE, NULL);
+  filter->ear_l_profile = g_strconcat (haar_path, DEFAULT_EAR_L_PROFILE, NULL);
   filter->display = TRUE;
   filter->face_detected = FALSE;
   filter->scale_factor = DEFAULT_SCALE_FACTOR;
@@ -390,6 +400,10 @@ gst_face_detect_init (GstFaceDetect * filter)
       gst_face_detect_load_profile (filter, filter->mouth_profile);
   filter->cvEyesDetect =
       gst_face_detect_load_profile (filter, filter->eyes_profile);
+  filter->cvEarLDetect =
+      gst_face_detect_load_profile (filter, filter->ear_l_profile);
+  filter->cvEarRDetect =
+      gst_face_detect_load_profile (filter, filter->ear_r_profile);
 
   gst_opencv_video_filter_set_in_place (GST_OPENCV_VIDEO_FILTER_CAST (filter),
       TRUE);
@@ -593,6 +607,8 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
     vector < Rect > mouth;
     vector < Rect > nose;
     vector < Rect > eyes;
+    vector < Rect > r_ears;
+    vector < Rect > l_ears;
     gboolean post_msg = FALSE;
 
     cvtColor (img, filter->cvGray, COLOR_RGB2GRAY);
@@ -646,8 +662,10 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
       guint rnx = 0, rny = 0, rnw, rnh;
       guint rmx = 0, rmy = 0, rmw, rmh;
       guint rex = 0, rey = 0, rew, reh;
+      guint relx = 0, rely = 0, relw, relh;
+      guint rerx = 0, rery = 0, rerw, rerh;
       guint rhh = r.height / 2;
-      gboolean have_nose, have_mouth, have_eyes;
+      gboolean have_nose, have_mouth, have_eyes, have_l_ears, have_r_ears;
 
       /* detect face features */
 
@@ -687,11 +705,35 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
         have_eyes = FALSE;
       }
 
+      if (filter->cvEarLDetect) {
+        relx = r.x;
+        rely = r.y;
+        relw = r.width;
+        relh = rhh;
+        gst_face_detect_run_detector (filter, filter->cvEarLDetect, mw, mh,
+            Rect (relx, rely, relw, relh), l_ears);
+        have_l_ears = !l_ears.empty ();
+      } else {
+        have_l_ears = FALSE;
+      }
+
+      if (filter->cvEarRDetect) {
+        rerx = r.x;
+        rery = r.y;
+        rerw = r.width;
+        rerh = rhh;
+        gst_face_detect_run_detector (filter, filter->cvEarRDetect, mw, mh,
+            Rect (rerx, rery, rerw, rerh), r_ears);
+        have_r_ears = !r_ears.empty ();
+      } else {
+        have_r_ears = FALSE;
+      }
+
       GST_LOG_OBJECT (filter,
           "%2d/%2" G_GSIZE_FORMAT
-          ": x,y = %4u,%4u: w.h = %4u,%4u : features(e,n,m) = %d,%d,%d", i,
-          faces.size (), r.x, r.y, r.width, r.height, have_eyes, have_nose,
-          have_mouth);
+          ": x,y = %4u,%4u: w.h = %4u,%4u : features(e,n,m) = %d,%d,%d,%d,%d",
+          i, faces.size (), r.x, r.y, r.width, r.height, have_eyes, have_nose,
+          have_mouth, have_r_ears, have_l_ears);
       if (post_msg) {
         s = gst_structure_new ("face",
             "x", G_TYPE_UINT, r.x,
@@ -704,6 +746,10 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
           structure_and_message (mouth, "mouth", rmx, rmy, filter, s);
         if (have_eyes)
           structure_and_message (eyes, "eyes", rex, rey, filter, s);
+        if (have_l_ears)
+          structure_and_message (l_ears, "left ears", relx, rely, filter, s);
+        if (have_r_ears)
+          structure_and_message (r_ears, "right ears", rerx, rery, filter, s);
 
         g_value_init (&facedata, GST_TYPE_STRUCTURE);
         g_value_take_boxed (&facedata, s);
@@ -761,6 +807,30 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
           axes.height = h;
           ellipse (img, center, axes, 0, 0, 360, Scalar (cr, cg, cb), 1, 8, 0);
         }
+        if (have_r_ears) {
+          Rect sr = r_ears[0];
+
+          w = sr.width / 2;
+          h = sr.height / 2;
+          center.x = cvRound ((rerx + sr.x + w));
+          center.y = cvRound ((rery + sr.y + h));
+          axes.width = w * 1.25; /* tweak for ears form */
+          axes.height = h;
+          ellipse (mtxOrg, center, axes, 0, 0, 360, Scalar (cr, cg, cb), 1, 8,
+              0);
+        }
+        if (have_l_ears) {
+          Rect sr = l_ears[0];
+
+          w = sr.width / 2;
+          h = sr.height / 2;
+          center.x = cvRound ((relx + sr.x + w));
+          center.y = cvRound ((rely + sr.y + h));
+          axes.width = w * 1.25; /* tweak for ears form */
+          axes.height = h;
+          ellipse (mtxOrg, center, axes, 0, 0, 360, Scalar (cr, cg, cb), 1, 8,
+              0);
+        }
       }
       gst_buffer_add_video_region_of_interest_meta (buf, "face",
           (guint) r.x, (guint) r.y, (guint) r.width, (guint) r.height);
diff --git a/ext/opencv/gstfacedetect.h b/ext/opencv/gstfacedetect.h
index e718844ce..68c159dd7 100644
--- a/ext/opencv/gstfacedetect.h
+++ b/ext/opencv/gstfacedetect.h
@@ -95,6 +95,8 @@ struct _GstFaceDetect
   gchar *nose_profile;
   gchar *mouth_profile;
   gchar *eyes_profile;
+  gchar *ear_l_profile;
+  gchar *ear_r_profile;
   gdouble scale_factor;
   gint min_neighbors;
   gint flags;
@@ -108,6 +110,8 @@ struct _GstFaceDetect
   cv::CascadeClassifier *cvNoseDetect;
   cv::CascadeClassifier *cvMouthDetect;
   cv::CascadeClassifier *cvEyesDetect;
+  cv::CascadeClassifier *cvEarLDetect;
+  cv::CascadeClassifier *cvEarRDetect;
 };
 
 struct _GstFaceDetectClass