def test_segm_supervised(self):
img = d_spl.load_sample_image(d_spl.IMAGE_DROSOPHILA_OVARY_2D)[:, :, 0]
img = imresize(img, (256, 256))
annot = d_spl.load_sample_image(d_spl.ANNOT_DROSOPHILA_OVARY_2D)
annot = imresize(annot, (256, 256), interp='nearest')
img2 = d_spl.load_sample_image(d_spl.IMAGE_DROSOPHILA_OVARY_2D)[:, :,
0]
img2 = imresize(img2, (256, 256))
path_dir = os.path.join(PATH_OUTPUT, 'test_segm_supervised_gc')
if not os.path.exists(path_dir):
os.mkdir(path_dir)
sp_size = 10
tp_edge = ['model', 'const']
list_regul = [0, 1, 10]
dict_imgs = dict()
name = 'fig-img%i_regul-%.2f_edge-%s.png'
classif, _, _, _ = pipelines.train_classif_color2d_slic_features(
[img], [annot], sp_size, dict_features=FEATURES_TEXTURE)
_ = pipelines.segment_color2d_slic_features_classif_graphcut(
img,
classif,
sp_size=sp_size,
gc_regul=0.,
dict_features=FEATURES_TEXTURE,
dict_debug_imgs=dict_imgs)
show_segm_debugs_2d(dict_imgs, path_dir, name % (1, 0, '_debug'))
for edge in tp_edge:
dict_imgs = dict()
for regul in list_regul:
seg = pipelines.segment_color2d_slic_features_classif_graphcut(
img,
classif,
sp_size=sp_size,
gc_regul=regul,
gc_edge_type=edge,
dict_features=FEATURES_TEXTURE)
show_segm_results_2d(img, seg, path_dir,
name % (1, regul, edge))
seg = pipelines.segment_color2d_slic_features_classif_graphcut(
img2,
classif,
sp_size=sp_size,
gc_regul=regul,
gc_edge_type=edge,
dict_features=FEATURES_TEXTURE,
dict_debug_imgs=dict_imgs)
show_segm_results_2d(img2, seg, path_dir,
name % (2, regul, edge))
show_segm_debugs_2d(dict_imgs, path_dir,
name % (2, regul, edge))
dict_imgs = None
def test_segm_gmm_gc_stars(self):
img = d_spl.load_sample_image(d_spl.IMAGE_STAR_2)
logging.debug('dimension: {}'.format(img.shape))
params = dict(nb_classes=3,
sp_regul=0.2,
sp_size=25,
dict_features={'color': ['mean', 'std']})
run_segm2d_gmm_gc(img, 'test_segm_gmm_gc_stars', dict_params=params)
def test_segm_gmm_gc_disc(self):
img = d_spl.load_sample_image(d_spl.IMAGE_DROSOPHILA_DISC)
img = imresize(img, (512, 512))
params = dict(sp_regul=0.2, sp_size=15, pca_coef=0.98)
run_segm2d_gmm_gc(img,
'test_segm_gmm_gc_disc',
types_edge=['model_l2'],
list_regul=[0, 1, 5],
dict_params=params)
def test_segm_gmm_gc_histo(self):
img = d_spl.load_sample_image(d_spl.IMAGE_HISTOL_FLAGSHIP)
img = imresize(img, (512, 512))
params = dict(sp_regul=0.15, sp_size=15, pca_coef=0.98)
run_segm2d_gmm_gc(img,
'test_segm_gmm_gc_histology',
types_edge=['model'],
list_regul=[0, 1, 5],
dict_params=params)
def test_segm_gmm_gc_langer(self):
img = d_spl.load_sample_image(d_spl.IMAGE_LANGER_ISLET)
img = imresize(img, (512, 512))
params = dict(clr_space='hsv', sp_regul=0.15, sp_size=5)
run_segm2d_gmm_gc(img,
'test_segm_gmm_gc_langer',
types_edge=['model_lT'],
list_regul=[0, 1],
dict_params=params)
def test_segm_gmm_gc_ovary_2d(self):
img = d_spl.load_sample_image(d_spl.IMAGE_DROSOPHILA_OVARY_2D)[:, :, 0]
img = imresize(img, (512, 512))
# img = np.rollaxis(np.tile(img[:, :, 0], (3, 1, 1)), 0, 3)
params = dict(nb_classes=4,
pca_coef=0.95,
sp_regul=0.3,
sp_size=10,
dict_features=seg_fts.FEATURES_SET_TEXTURE_SHORT)
run_segm2d_gmm_gc(img,
'test_segm_gmm_gc_ovary_2d',
list_regul=[0, 2, 10],
dict_params=params)
def test_segm_gmm_gc_stars(self):
img = d_spl.load_sample_image(d_spl.IMAGE_STAR_2)
logging.debug('dimension: {}'.format(img.shape))
params = {
'nb_classes': 3,
'pca_coef': 0.98,
'sp_regul': 0.2,
'sp_size': 25,
'dict_features': {
'color': ['mean', 'std']
}
}
run_segm2d_gmm_gc(img, 'test_segm_gmm_gc_stars', dict_params=params)
def test_count_transitions_segment(self):
img = d_spl.load_sample_image(d_spl.IMAGE_DROSOPHILA_OVARY_2D)[:, :, 0]
annot = d_spl.load_sample_image(d_spl.ANNOT_DROSOPHILA_OVARY_2D)
annot = annot.astype(int)
slic = seg_spx.segment_slic_img2d(img, sp_size=15, rltv_compact=0.2)
label_hist = seg_lb.histogram_regions_labels_norm(slic, annot)
labels = np.argmax(label_hist, axis=1)
trans = seg_gc.count_label_transitions_connected_segments({'a': slic},
{'a': labels})
path_csv = os.path.join(PATH_OUTPUT, 'labels_transitions.csv')
pd.DataFrame(trans).to_csv(path_csv)
gc_regul = seg_gc.compute_pairwise_cost_from_transitions(trans, 10.)
np.random.seed(0)
features = np.tile(labels, (5, 1)).T.astype(float)
features += np.random.random(features.shape) - 0.5
gmm = seg_gc.estim_class_model_gmm(features, 4)
proba = gmm.predict_proba(features)
seg_gc.segment_graph_cut_general(slic, proba, gc_regul)
def test_segm_gmm_gc_ovary_2d(self):
img = d_spl.load_sample_image(d_spl.IMAGE_DROSOPHILA_OVARY_2D)[:, :, 0]
img = imresize(img, (512, 512))
# img = np.rollaxis(np.tile(img[:, :, 0], (3, 1, 1)), 0, 3)
params = {
'nb_classes': 4,
'pca_coef': 0.95,
'sp_regul': 0.3,
'sp_size': 10,
'dict_features': seg_fts.FEATURES_SET_TEXTURE
}
run_segm2d_gmm_gc(img,
'test_segm_gmm_gc_ovary_2d',
list_regul=[0, 2, 10],
dict_params=params)
def test_show_image_features_clr2d(self):
img = d_spl.load_sample_image(d_spl.IMAGE_LENNA)
img = transform.resize(img, (128, 128))
slic = seg_spx.segment_slic_img2d(img, sp_size=10, rltv_compact=0.2)
features, names = seg_fts.compute_selected_features_color2d(
img, slic, seg_fts.FEATURES_SET_ALL)
path_dir = os.path.join(PATH_OUTPUT, 'test_image_rgb2d_features')
if not os.path.exists(path_dir):
os.mkdir(path_dir)
for i in range(features.shape[1]):
fts = features[:, i]
im_fts = fts[slic]
plt.imsave(os.path.join(path_dir, names[i] + '.png'), im_fts)
def test_segm_gmm_gc_objects(self):
img = d_spl.load_sample_image(d_spl.IMAGE_OBJECTS)
img = imresize(img, (256, 256))
logging.debug('dimension: {}'.format(img.shape))
params = {
'nb_classes': 4,
'sp_size': 20,
'sp_regul': 0.2,
'dict_features': {
'color': ['mean']
}
}
run_segm2d_gmm_gc(img,
'test_segm_gmm_gc_objects',
list_regul=[0, 1, 5, 10],
dict_params=params)
def test_segm_gmm_gc_objects(self):
img = d_spl.load_sample_image(d_spl.IMAGE_OBJECTS)
img = imresize(img, (256, 256))
logging.debug('dimension: {}'.format(img.shape))
dict_imgs = dict()
path_dir = os.path.join(PATH_OUTPUT, 'test_segm_gmm_gc_objects')
if not os.path.isdir(path_dir):
os.mkdir(path_dir)
seg = pipelines.pipe_color2d_slic_features_gmm_graphcut(
img,
nb_classes=4,
sp_size=20,
sp_regul=0.2,
dict_features={'color': ['mean']},
gc_regul=1.,
gc_edge_type='model',
dict_debug_imgs=dict_imgs)
show_segm_debugs_2d(dict_imgs, path_dir,
'fig_regul-%.2f_edge-%s_debug.png' % (1., 'model'))
show_segm_results_2d(img, seg, path_dir,
'fig_regul-%.2f_edge-%s.png' % (1., 'model'))
def setUp(self):
self.seg2d = d_spl.sample_segment_vertical_2d()
self.seg3d = d_spl.sample_segment_vertical_3d()
self.img = d_spl.load_sample_image(d_spl.IMAGE_LENNA)
