opencv: facedetect: DO NOT UPSTREAM: Rebased mods to facedetect

2 files changed, 169 insertions, 23 deletions

diff --git a/ext/opencv/gstfacedetect.cpp b/ext/opencv/gstfacedetect.cpp
index 503b9e70f..05227faa2 100644
--- a/ext/opencv/gstfacedetect.cpp
+++ b/ext/opencv/gstfacedetect.cpp
@@ -91,10 +91,12 @@ GST_DEBUG_CATEGORY_STATIC (gst_face_detect_debug);
 #define HAAR_CASCADES_DIR OPENCV_PREFIX G_DIR_SEPARATOR_S "share" \
     G_DIR_SEPARATOR_S OPENCV_PATH_NAME G_DIR_SEPARATOR_S "haarcascades" \
     G_DIR_SEPARATOR_S
-#define DEFAULT_FACE_PROFILE HAAR_CASCADES_DIR "haarcascade_frontalface_default.xml"
-#define DEFAULT_NOSE_PROFILE HAAR_CASCADES_DIR "haarcascade_mcs_nose.xml"
-#define DEFAULT_MOUTH_PROFILE HAAR_CASCADES_DIR "haarcascade_mcs_mouth.xml"
-#define DEFAULT_EYES_PROFILE HAAR_CASCADES_DIR "haarcascade_mcs_eyepair_small.xml"
+#define DEFAULT_FACE_PROFILE "haarcascade_frontalface_default.xml"
+#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
 #define DEFAULT_FLAGS CV_HAAR_DO_CANNY_PRUNING
 #define DEFAULT_MIN_NEIGHBORS 3
@@ -127,6 +129,12 @@ enum
   PROP_MIN_STDDEV
 };
 
+enum
+{
+  DIRECTION_INCONCLUSIVE = 0,
+  DIRECTION_RIGHT,
+  DIRECTION_LEFT
+};
 
 /*
  * GstOpencvFaceDetectFlags:
@@ -252,6 +260,8 @@ 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);
@@ -261,6 +271,10 @@ 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);
 }
@@ -291,22 +305,22 @@ gst_face_detect_class_init (GstFaceDetectClass * klass)
   g_object_class_install_property (gobject_class, PROP_FACE_PROFILE,
       g_param_spec_string ("profile", "Face profile",
           "Location of Haar cascade file to use for face detection",
-          DEFAULT_FACE_PROFILE,
+          HAAR_CASCADES_DIR DEFAULT_FACE_PROFILE,
           (GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
   g_object_class_install_property (gobject_class, PROP_NOSE_PROFILE,
       g_param_spec_string ("nose-profile", "Nose profile",
           "Location of Haar cascade file to use for nose detection",
-          DEFAULT_NOSE_PROFILE,
+          HAAR_CASCADES_DIR DEFAULT_NOSE_PROFILE,
           (GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
   g_object_class_install_property (gobject_class, PROP_MOUTH_PROFILE,
       g_param_spec_string ("mouth-profile", "Mouth profile",
           "Location of Haar cascade file to use for mouth detection",
-          DEFAULT_MOUTH_PROFILE,
+          HAAR_CASCADES_DIR DEFAULT_MOUTH_PROFILE,
           (GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
   g_object_class_install_property (gobject_class, PROP_EYES_PROFILE,
       g_param_spec_string ("eyes-profile", "Eyes profile",
           "Location of Haar cascade file to use for eye-pair detection",
-          DEFAULT_EYES_PROFILE,
+          HAAR_CASCADES_DIR DEFAULT_EYES_PROFILE,
           (GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
 
   g_object_class_install_property (gobject_class, PROP_FLAGS,
@@ -363,10 +377,17 @@ gst_face_detect_class_init (GstFaceDetectClass * klass)
 static void
 gst_face_detect_init (GstFaceDetect * filter)
 {
-  filter->face_profile = g_strdup (DEFAULT_FACE_PROFILE);
-  filter->nose_profile = g_strdup (DEFAULT_NOSE_PROFILE);
-  filter->mouth_profile = g_strdup (DEFAULT_MOUTH_PROFILE);
-  filter->eyes_profile = g_strdup (DEFAULT_EYES_PROFILE);
+
+  const gchar *haar_path;
+  haar_path = g_getenv ("GST_HAAR_CASCADES_PATH")?
+      g_getenv ("GST_HAAR_CASCADES_PATH") : HAAR_CASCADES_DIR;
+
+  filter->face_profile = g_strconcat (haar_path, DEFAULT_FACE_PROFILE, NULL);
+  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;
@@ -383,6 +404,10 @@ 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);
@@ -572,6 +597,37 @@ gst_face_detect_run_detector (GstFaceDetect * filter,
   }
 }
 
+/**
+ * Delta percentaje after wich position is considered off-center
+ */
+#define LINEAR_DAMPING_FACTOR 5
+
+/**
+ * @gst_face_detect_get_leaning_side
+ *
+ * Figures out what side of a reference a certain cuantity falls into
+ * @pos: Object x coordinate
+ * @ref_pos: Reference x coordinate
+ * @side: result. DIRECTION_RIGHT or DIRECTION_LEFT. Untouched if inconclusive
+ * @displacement: Magnitude of min to max ramp
+ */
+static inline void
+gst_face_detect_get_leaning_side (guint *side, gint reference, gint position,
+    guint displacement) {
+
+  g_assert (side);
+
+  displacement =  displacement * LINEAR_DAMPING_FACTOR / 100;
+
+  if ((guint)ABS (position - reference) <= displacement)
+    return;
+
+  if (position > reference)
+    *side = DIRECTION_LEFT;
+  else if (position < reference)
+    *side = DIRECTION_RIGHT;
+}
+
 /*
  * Performs the face detection
  */
@@ -590,6 +646,8 @@ 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;
 
     Mat mtxOrg (cv::cvarrToMat (img));
@@ -644,8 +702,10 @@ 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 */
 
@@ -685,11 +745,35 @@ 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,
@@ -702,6 +786,10 @@ 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);
@@ -711,20 +799,21 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
       }
 
       if (filter->display) {
-        CvPoint center;
-        Size axes;
+        CvPoint f_center, center;
+        Size f_axes, axes;
         gdouble w, h;
+        guint aim = DIRECTION_INCONCLUSIVE;
+
         gint cb = 255 - ((i & 3) << 7);
         gint cg = 255 - ((i & 12) << 5);
         gint cr = 255 - ((i & 48) << 3);
 
         w = r.width / 2;
         h = r.height / 2;
-        center.x = cvRound ((r.x + w));
-        center.y = cvRound ((r.y + h));
-        axes.width = w;
-        axes.height = h * 1.25; /* tweak for face form */
-        ellipse (mtxOrg, center, axes, 0, 0, 360, Scalar (cr, cg, cb), 3, 8, 0);
+        f_center.x = cvRound ((r.x + w));
+        f_center.y = cvRound ((r.y + h));
+        f_axes.width = w;
+        f_axes.height = h * 1.25; /* tweak for face form */
 
         if (have_nose) {
           Rect sr = nose[0];
@@ -733,11 +822,17 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
           h = sr.height / 2;
           center.x = cvRound ((rnx + sr.x + w));
           center.y = cvRound ((rny + sr.y + h));
+
+          /* left/right toggle */
+          gst_face_detect_get_leaning_side (&aim, f_center.x, center.x,
+              r.width);
+
           axes.width = w;
           axes.height = h * 1.25;       /* tweak for nose form */
           ellipse (mtxOrg, center, axes, 0, 0, 360, Scalar (cr, cg, cb), 1, 8,
               0);
         }
+
         if (have_mouth) {
           Rect sr = mouth[0];
 
@@ -745,11 +840,34 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
           h = sr.height / 2;
           center.x = cvRound ((rmx + sr.x + w));
           center.y = cvRound ((rmy + sr.y + h));
+
+          if (!aim) {
+            gst_face_detect_get_leaning_side (&aim, f_center.x, center.x,
+                r.width);
+          }
+
           axes.width = w * 1.5; /* tweak for mouth form */
           axes.height = h;
           ellipse (mtxOrg, center, axes, 0, 0, 360, Scalar (cr, cg, cb), 1, 8,
               0);
         }
+
+        /* Face ellipse */
+        switch (aim) {
+          case DIRECTION_RIGHT:
+            ellipse (mtxOrg, f_center, f_axes, 0, 0, 360, Scalar (cr, 0, 0),
+                3, 8, 0);
+            break;
+          case DIRECTION_LEFT:
+            ellipse (mtxOrg, f_center, f_axes, 0, 0, 360, Scalar (0, 0, cb),
+                3, 8, 0);
+            break;
+          case DIRECTION_INCONCLUSIVE:
+          default:
+            ellipse (mtxOrg, f_center, f_axes, 0, 0, 360, Scalar (cr, cg, cb),
+                3, 8, 0);
+        }
+
         if (have_eyes) {
           Rect sr = eyes[0];
 
@@ -762,6 +880,30 @@ gst_face_detect_transform_ip (GstOpencvVideoFilter * base, GstBuffer * buf,
           ellipse (mtxOrg, 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 03cf81c2e..81477d8fc 100644
--- a/ext/opencv/gstfacedetect.h
+++ b/ext/opencv/gstfacedetect.h
@@ -96,6 +96,8 @@ 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;
@@ -109,6 +111,8 @@ struct _GstFaceDetect
   cv::CascadeClassifier *cvNoseDetect;
   cv::CascadeClassifier *cvMouthDetect;
   cv::CascadeClassifier *cvEyesDetect;
+  cv::CascadeClassifier *cvEarLDetect;
+  cv::CascadeClassifier *cvEarRDetect;
 };
 
 struct _GstFaceDetectClass