def visual_coannotation(lnds_user, lnds_refs, path_dataset, path_user,
img_name, path_visu):
""" visualise the co-annotation
show consensus annotation and use annotation
:param DF lnds_user: loaded DataFrame
:param DF lnds_refs: loaded DataFrame
:param path_dataset: path to the image dataset (root)
:param name_set: name of the image set
:param name_user_scale: annotation folder containing user name and used scale
:param img_name: particular image/annotation/stain name
:param str|None path_visu: path to output
:return str: figure path
"""
name_set, name_user_scale = path_user.split(os.sep)[-2:]
user, scale = parse_path_user_scale(name_user_scale)
folder_scale = TEMPLATE_FOLDER_SCALE % scale
image = None
if path_dataset is not None and os.path.isdir(path_dataset):
path_dir = os.path.join(path_dataset, name_set, folder_scale)
paths_image = find_images(path_dir, img_name)
image = load_image(paths_image[0]) if paths_image else None
lnds_user = lnds_user * (scale / 100.)
lnds_refs = lnds_refs * (scale / 100.)
fig = figure_image_landmarks(lnds_refs, image, lnds_user, lnds2_name=user)
fig_name = NAME_FIGURE_COANNOT % img_name
path_dir = os.path.join(path_visu, name_set, name_user_scale)
create_folder_path(path_dir)
path_fig = os.path.join(path_dir, fig_name)
fig.savefig(path_fig)
plt.close(fig)
return path_fig
def load_image_landmarks(lnds_img_pair):
""" load image and related landmarks
:param (str, str) lnds_img_pair: tuple with paths
:return (str, str, ndarray, ndarray): image folder and name, landmarks and image
"""
p_lnds, p_img = lnds_img_pair
name = os.path.splitext(os.path.basename(p_img))[0]
lnd = pd.read_csv(p_lnds)
img = load_image(p_img)
folder = os.path.basename(os.path.dirname(p_lnds))
return folder, name, lnd, img
def export_visual_set_scale(d_paths):
""" export, visualise given set in particular scale
:param {str: str} d_paths: dictionary with path patterns
:return int: number of processed items
"""
list_lnds = sorted(glob.glob(os.path.join(d_paths['landmarks'], '*.csv')))
list_lnds_imgs = []
# fined relevant images to the given landmarks
for p_lnds in list_lnds:
name_ = os.path.splitext(os.path.basename(p_lnds))[0]
p_imgs = find_images(d_paths['images'], name_)
if p_imgs:
list_lnds_imgs.append((p_lnds, sorted(p_imgs)[0]))
# if there are no images or landmarks, skip it...
if not list_lnds_imgs:
logging.debug('no image-landmarks to show...')
return 0
# create the output folder
if not os.path.isdir(d_paths['output']):
os.makedirs(d_paths['output'])
# draw and export image-landmarks
for p_lnds, p_img in list_lnds_imgs:
name_ = os.path.splitext(os.path.basename(p_img))[0]
img = load_image(p_img)
if img is None:
continue
fig = figure_image_landmarks(pd.read_csv(p_lnds), img)
fig.savefig(os.path.join(d_paths['output'], name_ + '.pdf'))
plt.close(fig)
# draw and export PAIRS of image-landmarks
path_pairs = [(p1, p2) for i, p1 in enumerate(list_lnds_imgs)
for p2 in list_lnds_imgs[i + 1:]]
for p1, p2 in path_pairs:
export_visual_pairs(p1, p2, d_paths['output'])
return len(list_lnds_imgs)