Dlib should not use multiscale checking, made it faster

Dockerfile

 FROM smvanveen/computer-vision:20161109143812
 RUN git clone https://github.com/davisking/dlib.git
 RUN (cd dlib; python setup.py install --yes USE_AVX_INSTRUCTIONS)
+RUN apt-get install graphviz -y
+COPY requirements.txt /tmp
+RUN pip install -r /tmp/requirements.txt
 WORKDIR /src
-
-#COPY requirements.txt /tmp
-#RUN python --version

requirements.txt

-cycler==0.10.0
-matplotlib==1.5.3
-pyparsing==2.1.2
-python-dateutil==2.5.3
-pytz==2016.4
-requests==2.10.0
-six==1.10.0
-tornado==4.4.2
+pycallgraph==1.0.1

src/aam.py

@@ -199,7 +199,7 @@ def build_shape_feature_vectors(files, get_points, flattened=False):
     return points
 
 
-def sample_from_triangles(src, points2d_src, points2d_dst, triangles, dst):
+def piecewise_transform(src, points2d_src, points2d_dst, triangles, dst):
     """
     Get pixels from within the  triangles [[p1, p2, p3]_0, .. [p1, p2, p3]_n].
 
@@ -259,7 +259,7 @@ def build_texture_feature_vectors(
         # empty colored image
         dst = np.full((image.shape[0], image.shape[1], 3), fill_value=0, dtype=np.uint8)
 
-        sample_from_triangles(
+        piecewise_transform(
             image,
             Points.get_scaled_points(image.shape),
             mean_points.get_scaled_points(image.shape),

src/landmarks.py

@@ -51,7 +51,8 @@ class Detector():
         # The 1 in the second argument indicates that we should upsample the
         # image 1 time.  This will make everything bigger and allow us to
         # detect more faces.
-        return self.detector(image, 1)
+        # 0 = no upsample
+        return self.detector(image, 0)
 
     def detect_shape(self, image):
         detections = self.detect_faces(image)

src/reconstruction/reconstruction.py

@@ -125,10 +125,11 @@ def reconstruct_texture(src_image, dst_image, texture_model,
     points2d_dst[:, 0] -= offset_x
     points2d_dst[:, 1] -= offset_y
 
-    points2d_src = points2d_src * 1.1
+    # don't know why this was, removing still works, keeping it for a while.
+    #points2d_src = points2d_src * 1.1
 
     # get the texture from the rectangles.
-    aam.sample_from_triangles(
+    aam.piecewise_transform(
         r_texture,
         points2d_dst,  # turn src and dst around
         points2d_output,  # turn src and dst around
@@ -141,7 +142,7 @@ def reconstruct_shape_texture(dataset_module, shape_model, texture_model,
                               image_filename, shape_components,
                               shape_eigenvalues_multiplier=[],
                               image_as_background=False):
-    """Reconstructs shape and texture"""
+    """Performs reconstruction for shape and texture."""
     input_points = dataset_module.factory(filename=image_filename)
     input_image = input_points.get_image()