def build(self, images, gt_shapes, boxes):
self.mean_shape = util.centered_mean_shape(gt_shapes)
self.n_landmarks = self.mean_shape.n_points
# Generate initial shapes with perturbations.
print_dynamic('Generating initial shapes')
shapes = np.array([util.fit_shape_to_box(self.mean_shape, box) for box in boxes])
print_dynamic('Perturbing initial estimates')
if self.n_perturbations > 1:
images, shapes, gt_shapes, boxes = util.perturb_shapes(images, shapes, gt_shapes, boxes,
self.n_perturbations, mode='mean_shape')
assert(len(boxes) == len(images))
assert(len(shapes) == len(images))
assert(len(gt_shapes) == len(images))
print('\nSize of augmented dataset: {} images.\n'.format(len(images)))
weak_regressors = []
for j in xrange(self.n_stages):
# Calculate normalized targets.
deltas = [gt_shapes[i].points - shapes[i].points for i in xrange(len(images))]
targets = np.array([util.transform_to_mean_shape(shapes[i], self.mean_shape).apply(deltas[i]).reshape((2*self.n_landmarks,))
for i in xrange(len(images))])
weak_regressor = self.weak_builder.build(images, targets, (shapes, self.mean_shape, j))
# Update current estimates of shapes.
for i in xrange(len(images)):
offset = weak_regressor.apply(images[i], shapes[i])
shapes[i].points += offset.points
weak_regressors.append(weak_regressor)
print("\nBuilt outer regressor {}\n".format(j))
return CascadedShapeRegressor(self.n_landmarks, weak_regressors, self.mean_shape)
def apply(self, image, shape):
mean_to_shape = util.transform_to_mean_shape(shape, self.mean_shape).pseudoinverse()
shape_indexed_features = self.feature_extractor.extract_features(image, shape, mean_to_shape)
res = PointCloud(np.zeros((self.n_landmarks, 2)), copy=False)
for r in self.regressors:
offset = r.apply(shape_indexed_features, self.extra)
res.points += offset.reshape((self.n_landmarks, 2))
return mean_to_shape.apply(res)
def apply(self, image, shape):
mean_to_shape = util.transform_to_mean_shape(shape, self.mean_shape).pseudoinverse()
#feat = self.feature_extractor.apply(image, shape)
#res = self.regression_matrix[feat == 1].sum(axis=0).reshape((self.n_landmarks, 2))
#return mean_to_shape.apply(PointCloud(res, copy=False))
indices = self.feature_extractor.get_indices(image, shape, mean_to_shape)
return mean_to_shape.apply(PointCloud(self.regression_matrix[indices].sum(axis=0).reshape((self.n_landmarks, 2)),
copy=False))
def apply(self, image, shape):
mean_to_shape = util.transform_to_mean_shape(shape, self.mean_shape).pseudoinverse()
#feat = self.feature_extractor.apply(image, shape)
#res = self.regression_matrix[feat == 1].sum(axis=0).reshape((self.n_landmarks, 2))
#return mean_to_shape.apply(PointCloud(res, copy=False))
indices = self.feature_extractor.get_indices(image, shape, mean_to_shape)
return mean_to_shape.apply(PointCloud(self.regression_matrix[indices].sum(axis=0).reshape((self.n_landmarks, 2)),
copy=False))
def apply(self, image, shape):
mean_to_shape = util.transform_to_mean_shape(
shape, self.mean_shape).pseudoinverse()
shape_indexed_features = self.feature_extractor.extract_features(
image, shape, mean_to_shape)
res = PointCloud(np.zeros((self.n_landmarks, 2)), copy=False)
for r in self.regressors:
offset = r.apply(shape_indexed_features, self.extra)
res.points += offset.reshape((self.n_landmarks, 2))
return mean_to_shape.apply(res)
def apply(self, img, shape):
n_landmarks = len(self.forests)
n_trees = len(self.forests[0].regressors)
n_leaves = len(self.forests[0].regressors[0].leaves)
local_binary_features = np.zeros(n_landmarks*n_trees*n_leaves)
mean_to_shape = util.transform_to_mean_shape(shape, self.mean_shape).pseudoinverse()
for landmark_i, f in enumerate(self.forests):
pixels = f.feature_extractor.extract_features(img, shape, mean_to_shape)
for tree_i, tree in enumerate(f.regressors):
leaf = tree.get_leaf_index(pixels)
local_binary_features[landmark_i*n_trees*n_leaves + tree_i*n_leaves + leaf] = 1
return local_binary_features
def build(self, images, gt_shapes, boxes):
self.mean_shape = util.centered_mean_shape(gt_shapes)
self.n_landmarks = self.mean_shape.n_points
# Generate initial shapes with perturbations.
print_dynamic('Generating initial shapes')
shapes = np.array(
[util.fit_shape_to_box(self.mean_shape, box) for box in boxes])
print_dynamic('Perturbing initial estimates')
if self.n_perturbations > 1:
images, shapes, gt_shapes, boxes = util.perturb_shapes(
images,
shapes,
gt_shapes,
boxes,
self.n_perturbations,
mode='mean_shape')
assert (len(boxes) == len(images))
assert (len(shapes) == len(images))
assert (len(gt_shapes) == len(images))
print('\nSize of augmented dataset: {} images.\n'.format(len(images)))
weak_regressors = []
for j in xrange(self.n_stages):
# Calculate normalized targets.
deltas = [
gt_shapes[i].points - shapes[i].points
for i in xrange(len(images))
]
targets = np.array([
util.transform_to_mean_shape(shapes[i], self.mean_shape).apply(
deltas[i]).reshape((2 * self.n_landmarks, ))
for i in xrange(len(images))
])
weak_regressor = self.weak_builder.build(
images, targets, (shapes, self.mean_shape, j))
# Update current estimates of shapes.
for i in xrange(len(images)):
offset = weak_regressor.apply(images[i], shapes[i])
shapes[i].points += offset.points
weak_regressors.append(weak_regressor)
print("\nBuilt outer regressor {}\n".format(j))
return CascadedShapeRegressor(self.n_landmarks, weak_regressors,
self.mean_shape)
def apply(self, img, shape):
n_landmarks = len(self.forests)
n_trees = len(self.forests[0].regressors)
n_leaves = len(self.forests[0].regressors[0].leaves)
local_binary_features = np.zeros(n_landmarks * n_trees * n_leaves)
mean_to_shape = util.transform_to_mean_shape(
shape, self.mean_shape).pseudoinverse()
for landmark_i, f in enumerate(self.forests):
pixels = f.feature_extractor.extract_features(
img, shape, mean_to_shape)
for tree_i, tree in enumerate(f.regressors):
leaf = tree.get_leaf_index(pixels)
local_binary_features[landmark_i * n_trees * n_leaves +
tree_i * n_leaves + leaf] = 1
return local_binary_features
def to_mean(self, shape):
return util.transform_to_mean_shape(shape, self.mean_shape)
def to_mean(self, shape):
return util.transform_to_mean_shape(shape, self.mean_shape)