Upgraded render_face_detections() and get_face_chip_details() to support 5

point face landmark models.

dlib/image_processing/render_face_detections.h

@@ -18,13 +18,23 @@ namespace dlib
         std::vector<image_window::overlay_line> lines;
         for (unsigned long i = 0; i < dets.size(); ++i)
         {
-            DLIB_CASSERT(dets[i].num_parts() == 68,
+            DLIB_CASSERT(dets[i].num_parts() == 68 || dets[i].num_parts() == 5,
                 "\t std::vector<image_window::overlay_line> render_face_detections()"
-                << "\n\t Invalid inputs were given to this function. "
+                << "\n\t You have to give either a 5 point or 68 point face landmarking output to this function. "
                 << "\n\t dets["<<i<<"].num_parts():  " << dets[i].num_parts() 
             );
 
             const full_object_detection& d = dets[i];
+
+            if (d.num_parts() == 5)
+            {
+                lines.push_back(image_window::overlay_line(d.part(0), d.part(1), color));
+                lines.push_back(image_window::overlay_line(d.part(1), d.part(4), color));
+                lines.push_back(image_window::overlay_line(d.part(4), d.part(3), color));
+                lines.push_back(image_window::overlay_line(d.part(3), d.part(2), color));
+            }
+            else
+            {
                 // Around Chin. Ear to Ear
                 for (unsigned long i = 1; i <= 16; ++i)
                     lines.push_back(image_window::overlay_line(d.part(i), d.part(i-1), color));
@@ -65,6 +75,7 @@ namespace dlib
                     lines.push_back(image_window::overlay_line(d.part(i), d.part(i-1), color));
                 lines.push_back(image_window::overlay_line(d.part(60), d.part(67), color));
             }
+        }
         return lines;
     }
 

dlib/image_processing/render_face_detections_abstract.h

@@ -18,14 +18,21 @@ namespace dlib
     /*!
         requires
             - for all valid i:
-                - dets[i].num_parts() == 68
+                - dets[i].num_parts() == 68 || dets[i].num_parts() == 5
         ensures
             - Interprets the given objects as face detections with parts annotated using
-              the iBUG face landmark scheme.  We then return a set of overlay lines that
-              will draw the objects onto the screen in a way that properly draws the
-              outline of the face features defined by the part locations.
+              either the iBUG face landmark scheme or a 5 point face annotation.  We then
+              return a set of overlay lines that will draw the objects onto the screen in a
+              way that properly draws the outline of the face features defined by the part
+              locations.
             - returns a vector with dets.size() elements, each containing the lines
               necessary to render a face detection from dets.
+            - The 5 point face annotation scheme is assumed to be:
+                - det part 0 == left eye corner, outside part of eye.
+                - det part 1 == left eye corner, inside part of eye.
+                - det part 2 == right eye corner, outside part of eye.
+                - det part 3 == right eye corner, inside part of eye.
+                - det part 4 == immediately under the nose, right at the top of the philtrum.
     !*/
 
 // ----------------------------------------------------------------------------------------
@@ -36,7 +43,7 @@ namespace dlib
     );
     /*!
         requires
-            - det.num_parts() == 68
+            - det.num_parts() == 68 || det.num_parts() == 5
         ensures
             - This function is identical to the above render_face_detections() routine
               except that it takes just a single full_object_detection instead of a

dlib/image_transforms/interpolation.h

@@ -1991,9 +1991,9 @@ namespace dlib
         const double padding = 0.2
     )
     {
-        DLIB_CASSERT(det.num_parts() == 68,
+        DLIB_CASSERT(det.num_parts() == 68 || det.num_parts() == 5,
             "\t chip_details get_face_chip_details()"
-            << "\n\t You must give a detection with exactly 68 parts in it."
+            << "\n\t You have to give either a 5 point or 68 point face landmarking output to this function. "
             << "\n\t det.num_parts(): " << det.num_parts()
         );
         DLIB_CASSERT(padding >= 0 && size > 0,
@@ -2003,6 +2003,40 @@ namespace dlib
             << "\n\t size:    " << size 
             );
 
+
+        std::vector<dpoint> from_points, to_points;
+        if (det.num_parts() == 5)
+        {
+            dpoint p0(0.8595674595992, 0.2134981538014);
+            dpoint p1(0.6460604764104, 0.2289674387677);
+            dpoint p2(0.1205750620789, 0.2137274526848);
+            dpoint p3(0.3340850613712, 0.2290642403242);
+            dpoint p4(0.4901123135679, 0.6277975316475);
+
+
+            p0 = (padding+p0)/(2*padding+1);
+            p1 = (padding+p1)/(2*padding+1);
+            p2 = (padding+p2)/(2*padding+1);
+            p3 = (padding+p3)/(2*padding+1);
+            p4 = (padding+p4)/(2*padding+1);
+
+            from_points.push_back(p0*size);
+            to_points.push_back(det.part(0));
+
+            from_points.push_back(p1*size);
+            to_points.push_back(det.part(1));
+
+            from_points.push_back(p2*size);
+            to_points.push_back(det.part(2));
+
+            from_points.push_back(p3*size);
+            to_points.push_back(det.part(3));
+
+            from_points.push_back(p4*size);
+            to_points.push_back(det.part(4));
+        }
+        else
+        {
             // Average positions of face points 17-67
             const double mean_face_shape_x[] = {
                 0.000213256, 0.0752622, 0.18113, 0.29077, 0.393397, 0.586856, 0.689483, 0.799124,
@@ -2025,7 +2059,6 @@ namespace dlib
 
             COMPILE_TIME_ASSERT(sizeof(mean_face_shape_x)/sizeof(double) == 68-17);
 
-        std::vector<dlib::vector<double,2> > from_points, to_points;
             for (unsigned long i = 17; i < det.num_parts(); ++i)
             {
                 // Ignore the lower lip
@@ -2035,12 +2068,20 @@ namespace dlib
                 if (17 <= i && i <= 26)
                     continue;
 
-            dlib::vector<double,2> p;
+                if (1+i != 37 && 
+                    1+i != 40 &&
+                    1+i != 43 &&
+                    1+i != 46 &&
+                    1+i != 34)
+                    continue;
+
+                dpoint p;
                 p.x() = (padding+mean_face_shape_x[i-17])/(2*padding+1);
                 p.y() = (padding+mean_face_shape_y[i-17])/(2*padding+1);
                 from_points.push_back(p*size);
                 to_points.push_back(det.part(i));
             }
+        }
 
         return chip_details(from_points, to_points, chip_dims(size,size));
     }

dlib/image_transforms/interpolation_abstract.h

@@ -1365,20 +1365,30 @@ namespace dlib
     );
     /*!
         requires
-            - det.num_parts() == 68
+            - det.num_parts() == 68 || det.num_parts() == 5
             - size > 0
             - padding >= 0
         ensures
             - This function assumes det contains a human face detection with face parts
               annotated using the annotation scheme from the iBUG 300-W face landmark
-              dataset.  Given these assumptions, it creates a chip_details object that will
-              extract a copy of the face that has been rotated upright, centered, and
-              scaled to a standard size when given to extract_image_chip(). 
+              dataset or a 5 point face annotation.  Given these assumptions, it creates a
+              chip_details object that will extract a copy of the face that has been
+              rotated upright, centered, and scaled to a standard size when given to
+              extract_image_chip(). 
+            - This function is specifically calibrated to work with one of these models:
+                - http://dlib.net/files/shape_predictor_5_face_landmarks.dat.bz2
+                - http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
             - The extracted chips will have size rows and columns in them.
             - if padding == 0 then the chip will be closely cropped around the face.
               Setting larger padding values will result a looser cropping.  In particular,
               a padding of 0.5 would double the width of the cropped area, a value of 1
               would triple it, and so forth.
+            - The 5 point face annotation scheme is assumed to be:
+                - det part 0 == left eye corner, outside part of eye.
+                - det part 1 == left eye corner, inside part of eye.
+                - det part 2 == right eye corner, outside part of eye.
+                - det part 3 == right eye corner, inside part of eye.
+                - det part 4 == immediately under the nose, right at the top of the philtrum.
     !*/
 
 // ----------------------------------------------------------------------------------------