def segment_graph_cut_general(segments,
proba,
image=None,
features=None,
gc_regul=1.,
edge_type='model',
edge_cost=1.,
debug_visual=None):
""" segment the image segmented via superpixels and estimated features
:param ndarray features: features sor each instance
:param ndarray segments: segmentation mapping each pixel into a class
:param ndarray image: image
:param ndarray proba: probabilities that each feature belongs to each class
:param str edge_type:
:param str edge_cost:
:param gc_regul: regularisation for GrphCut
:param {} debug_visual:
:return [int]: labelling by resulting classes
>>> np.random.seed(0)
>>> segments = np.array([[0] * 3 + [2] * 3 + [4] * 3 + [6] * 3 + [8] * 3,
... [1] * 3 + [3] * 3 + [5] * 3 + [7] * 3 + [9] * 3])
>>> proba = np.array([[0.1] * 6 + [0.9] * 4,
... [0.9] * 6 + [0.1] * 4], dtype=float).T
>>> proba += (0.5 - np.random.random(proba.shape)) * 0.2
>>> compute_unary_cost(proba)
array([[ 2.40531242, 0.15436155],
[ 2.53266106, 0.11538463],
[ 2.1604864 , 0.13831863],
[ 2.18495711, 0.19644636],
[ 4.60517019, 0.0797884 ],
[ 3.17833405, 0.11180231],
[ 0.12059702, 4.20769207],
[ 0.0143091 , 1.70059894],
[ 0.01005034, 3.39692559],
[ 0.16916609, 3.64975219]])
>>> segment_graph_cut_general(segments, proba, gc_regul=0., edge_type='')
array([1, 1, 1, 1, 1, 1, 0, 0, 0, 0], dtype=int32)
>>> labels = segment_graph_cut_general(segments, proba, gc_regul=1.,
... edge_type='spatial')
>>> labels[segments]
array([[1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0]], dtype=int32)
>>> slic = np.array([[0] * 4 + [1] * 6 + [2] * 4,
... [3] * 5 + [4] * 4 + [5] * 5])
>>> proba = np.array([[1] * 3 + [0] * 3, [0] * 3 + [1] * 3], dtype=float).T
>>> proba += np.random.random(proba.shape) / 2.
>>> np.argmax(proba, axis=1)
array([0, 0, 0, 1, 1, 1])
>>> debug_visual = dict()
>>> segment_graph_cut_general(slic, proba, gc_regul=0., edge_type='',
... debug_visual=debug_visual)
array([0, 0, 0, 1, 1, 1], dtype=int32)
>>> sorted(debug_visual.keys()) #doctest: +NORMALIZE_WHITESPACE
['edge_weights', 'edges', 'img_graph_edges', 'img_graph_segm',
'imgs_unary_cost', 'segments']
"""
logging.debug('convert variables and run GraphCut on created graph.')
edges, edge_weights = compute_edge_weights(segments, image, features,
proba, edge_type)
edge_weights *= edge_cost
logging.debug('graph edges weights %r', edge_weights.shape)
unary_cost = compute_unary_cost(proba)
logging.debug('graph unaries potentials: %r', unary_cost.shape)
pairwise_cost = compute_pairwise_cost(gc_regul, proba.shape)
logging.debug('graph pairwise coefs: \n%r', pairwise_cost)
if gc_regul <= 0:
logging.debug('gc_regul=%f so we use just argmax()', gc_regul)
graph_labels = np.argmin(unary_cost, axis=-1).astype(np.int32)
else:
# run GraphCut
logging.debug('perform GraphCut')
graph_labels = cut_general_graph(
edges,
edge_weights,
unary_cost,
pairwise_cost,
algorithm='expansion',
# down_weight_factor=np.abs(unary_cost).max(),
# init_labels=np.argmax(proba, axis=1),
n_iter=-1)
insert_gc_debug_images(debug_visual, segments, graph_labels,
compute_unary_cost(proba), edges, edge_weights)
return graph_labels
def segment_graph_cut_general(segments,
proba,
image=None,
features=None,
gc_regul=1.,
edge_type='model',
edge_cost=1.,
dict_debug_imgs=None):
""" segment the image segmented via superpixels and estimated features
:param ndarray features: features sor each instance
:param ndarray segments: segmentation mapping each pixel into a class
:param ndarray proba: probabilities that each feature belongs to each class
:param gc_regul: regularisation for GrphCut
:param {} dict_debug_imgs:
:return [int]: labelling by resulting classes
>>> np.random.seed(0)
>>> segments = np.array([[0] * 3 + [1] * 3 + [2] * 3 + [3] * 3 + [4] * 3,
... [5] * 3 + [6] * 3 + [7] * 3 + [8] * 3 + [9] * 3])
>>> proba = np.array([[0] * 6 + [1] * 4, [1] * 6 + [0] * 4], dtype=float).T
>>> proba += np.random.random(proba.shape) / 2.
>>> compute_unary_cost(proba)
array([[ 1.29314378, 0.30571452],
[ 1.19937775, 0.24093757],
[ 1.55198349, 0.27986187],
[ 1.51962643, 0.36872263],
[ 0.73016106, 0.17539828],
[ 0.9266883 , 0.23463524],
[ 0.24999756, 0.77046392],
[ 0.03490181, 3.13350924],
[ 0.01005844, 0.87632529],
[ 0.32864049, 0.83239528]])
>>> segment_graph_cut_general(segments, proba, gc_regul=0., edge_type='')
array([1, 1, 1, 1, 1, 1, 0, 0, 0, 0], dtype=int32)
>>> labels = segment_graph_cut_general(segments, proba, gc_regul=1.,
... edge_type='spatial')
>>> labels[segments]
array([[1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=int32)
>>> slic = np.array([[0] * 4 + [1] * 6 + [2] * 4,
... [3] * 5 + [4] * 4 + [5] * 5])
>>> proba = np.array([[1] * 3 + [0] * 3, [0] * 3 + [1] * 3], dtype=float).T
>>> proba += np.random.random(proba.shape) / 2.
>>> np.argmax(proba, axis=1)
array([0, 0, 0, 1, 1, 1])
>>> dict_debug_imgs = dict()
>>> segment_graph_cut_general(slic, proba, gc_regul=0., edge_type='',
... dict_debug_imgs=dict_debug_imgs)
array([0, 0, 0, 1, 1, 1], dtype=int32)
>>> sorted(dict_debug_imgs.keys()) #doctest: +NORMALIZE_WHITESPACE
['edge_weights', 'edges', 'img_graph_edges', 'img_graph_segm',
'imgs_unary_cost', 'segments']
"""
logging.debug('convert variables and run GraphCut on created graph.')
edges, edge_weights = compute_edge_weights(segments, image, features,
proba, edge_type)
edge_weights *= edge_cost
logging.debug('graph edges weights %s', repr(edge_weights.shape))
unary_cost = compute_unary_cost(proba)
logging.debug('graph unaries potentials: %s', repr(unary_cost.shape))
pairwise_cost = compute_pairwise_cost(gc_regul, proba.shape)
logging.debug('graph pairwise coefs: \n%s', repr(pairwise_cost))
labels = np.argmax(proba, axis=1)
# run GraphCut
logging.debug('perform GraphCut')
graph_labels = cut_general_graph(
edges,
edge_weights,
unary_cost,
pairwise_cost,
algorithm='expansion',
# down_weight_factor=np.abs(unary_cost).max()
init_labels=labels,
n_iter=9999)
insert_gc_debug_images(dict_debug_imgs, segments, graph_labels,
compute_unary_cost(proba), edges, edge_weights)
return graph_labels
