def evaluate_user(user_name,
path_annots,
path_out,
path_dataset=None,
visual=False):
""" evaluate single user statistic against consensus
:param str user_name: annotator name
:param str path_annots: path to the root of annotations
:param str path_out: path for statistic and visualisation
:param str path_dataset: path to image dataset
:param bool visual: visualise also outputs
:return int: processed items
"""
tissue_sets = list_sub_folders(path_annots)
path_visu = path_out if visual else None
stats = []
for p_set in tissue_sets:
paths = list_sub_folders(p_set, '*_scale-*pc')
user_names = [parse_path_user_scale(p)[0].lower() for p in paths]
paths_lnds_user = [
p for p, u in zip(paths, user_names) if u == user_name.lower()
]
paths_lnds_refs = [
p for p, u in zip(paths, user_names) if u != user_name.lower()
]
if not paths_lnds_user or not paths_lnds_refs:
continue
for path_user in paths_lnds_user:
stats += compute_statistic(path_user, paths_lnds_refs,
path_dataset, path_visu)
if not stats:
logging.warning('no statistic collected')
return 0
df_stats = pd.DataFrame(stats)
df_stats.set_index(['name_image_set', 'name_user', 'landmarks'],
inplace=True)
path_csv = os.path.join(path_out, 'STATISTIC_%s.csv' % user_name)
logging.debug('exporting CSV stat.: %s', path_csv)
df_stats.to_csv(path_csv)
df_stat_short = df_stats.describe().T[['count', 'mean', 'std', 'max']]
logging.info('USER: %s \n%s \n %s' % (user_name, '=' * 10, df_stat_short))
return len(df_stats)
def evaluate_user(user_name,
path_annots,
path_out,
path_dataset=None,
tp_consensus='mean',
visual=False):
""" evaluate single user statistic against consensus
:param str user_name: annotator name
:param str path_annots: path to the root of annotations
:param str tp_consensus: type of consensus landmarks
:param str path_out: path for statistic and visualisation
:param str path_dataset: path to image dataset
:param bool visual: visualise also outputs
:return int: processed items
"""
tissue_sets = list_sub_folders(path_annots)
path_visu = path_out if visual else None
stats = []
for p_set in tissue_sets:
paths = list_sub_folders(p_set, '*_scale-*pc')
user_names = [parse_path_user_scale(p)[0].lower() for p in paths]
paths_lnds_user = [
p for p, u in zip(paths, user_names) if u == user_name.lower()
]
paths_lnds_refs = [
p for p, u in zip(paths, user_names) if u != user_name.lower()
]
if not paths_lnds_user or not paths_lnds_refs:
continue
for path_user in paths_lnds_user:
stats += compute_statistic(path_user, paths_lnds_refs,
tp_consensus, path_dataset, path_visu)
if not stats:
logging.warning('no statistic collected')
return 0
df_stats = pd.DataFrame(stats)
return df_stats
def dataset_scale_landmarks(path_dataset, scales=SCALES, nb_jobs=NB_THREADS):
"""" scale whole dataset
:param str path_dataset:
:param [int] scales: selected output scales
:param nb_jobs: run parallel jobs
:return [int]:
"""
list_sets = list_sub_folders(path_dataset)
logging.info('Found sets: %i', len(list_sets))
_wrap_scale = partial(scale_set_landmarks, scales=scales)
counts = list(wrap_execute_parallel(
_wrap_scale, sorted(list_sets), nb_jobs=nb_jobs, desc='scaling sets'))
return counts
def dataset_generate_landmarks(path_annots, path_dataset,
nb_jobs=NB_THREADS):
""" generate consensus landmarks in full scale (100%)
:param str path_annots: path to folder with annotations
:param str path_dataset: output dataset path
:param nb_jobs: run parallel jobs
:return [int]:
"""
list_sets = list_sub_folders(path_annots)
logging.info('Found sets: %i', len(list_sets))
_wrap_lnds = partial(generate_consensus_landmarks, path_dataset=path_dataset)
counts = list(wrap_execute_parallel(
_wrap_lnds, sorted(list_sets), nb_jobs=nb_jobs, desc='consensus landmarks'))
return counts
def generate_consensus_landmarks(path_set, path_dataset):
""" generate consensus landmarks for a particular image/landmark set
:param str path_set: path to the set with annotations
:param str path_dataset: output dataset path (root)
:return {str: int}:
"""
path_annots = list_sub_folders(path_set, '*_scale-*pc')
logging.debug('>> found annotations: %i', len(path_annots))
dict_lnds, dict_lens = create_consensus_landmarks(path_annots)
path_scale = os.path.join(path_dataset, os.path.basename(path_set),
TEMPLATE_FOLDER_SCALE % 100)
create_folder_path(path_scale)
for name in dict_lnds:
dict_lnds[name].to_csv(os.path.join(path_scale, name))
return {os.path.basename(path_set): dict_lens}