def main(paths, nb_jobs=NB_THREADS):
logging.info('running...')
logging.info(tl_expt.string_dict(paths, desc='PATHS'))
if not os.path.exists(paths['output']):
assert os.path.isdir(os.path.dirname(paths['output'])), \
'missing folder: %s' % os.path.dirname(paths['output'])
os.mkdir(paths['output'])
paths_imgs = glob.glob(paths['images'])
logging.info('found %i images in dir "%s"', len(paths_imgs),
paths['images'])
warped_overlap = partial(perform_visu_overlap, paths=paths)
created = []
iterate = tl_expt.WrapExecuteSequence(warped_overlap,
paths_imgs,
nb_jobs=nb_jobs,
desc='overlapping')
for r in iterate:
created.append(r)
logging.info('matched and created %i overlaps', np.sum(created))
logging.info('DONE')
def main(paths, nb_jobs=NB_THREADS, segm_alpha=MIDDLE_ALPHA_OVERLAP):
logging.info('running...')
assert paths['segms'] != paths['output'], 'overwriting segmentation dir'
assert os.path.basename(paths['images']) != paths['output'], \
'overwriting image dir'
logging.info(tl_expt.string_dict(paths, desc='PATHS'))
if not os.path.exists(paths['output']):
assert os.path.isdir(os.path.dirname(paths['output'])), \
'missing folder: %s' % os.path.dirname(paths['output'])
os.mkdir(paths['output'])
paths_imgs = glob.glob(paths['images'])
logging.info('found %i images in dir "%s"', len(paths_imgs), paths['images'])
_warped_overlap = partial(perform_visu_overlap, paths=paths,
segm_alpha=segm_alpha)
created = []
iterate = tl_expt.WrapExecuteSequence(_warped_overlap, paths_imgs,
nb_jobs=nb_jobs, desc='overlapping')
for r in iterate:
created.append(r)
logging.info('matched and created %i overlaps', np.sum(created))
logging.info('DONE')
def main(params, debug_export=DEBUG_EXPORT):
""" the main entry point
:param {str: ...} params: segmentation parameters
:param bool debug_export: whether export visualisations
"""
logging.getLogger().setLevel(logging.DEBUG)
params = tl_expt.create_experiment_folder(
params, dir_name=NAME_EXPERIMENT, stamp_unique=EACH_UNIQUE_EXPERIMENT)
tl_expt.set_experiment_logger(params['path_exp'])
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
# tl_expt.create_subfolders(params['path_exp'], [FOLDER_IMAGE])
df_paths = pd.read_csv(params['path_list'], index_col=0)
logging.info('loaded %i items with columns: %s', len(df_paths),
repr(df_paths.columns.tolist()))
df_paths.dropna(how='any', inplace=True)
# create sub-folders if required
tl_expt.create_subfolders(params['path_exp'], ['input', 'simple'])
dict_segment = create_dict_segmentation(params, None, None, None, None)
tl_expt.create_subfolders(params['path_exp'], [n for n in dict_segment] +
[n + DIR_CENTRE_POSIX for n in dict_segment] +
[n + DIR_VISUAL_POSIX for n in dict_segment])
if debug_export:
list_dirs = [n + DIR_DEBUG_POSIX for n in dict_segment if 'rg2sp' in n]
tl_expt.create_subfolders(params['path_exp'], list_dirs)
_wrapper_segment = partial(image_segmentation, params=params)
iterate = tl_expt.WrapExecuteSequence(_wrapper_segment,
df_paths.iterrows(),
nb_jobs=params['nb_jobs'])
list(iterate)
def main(params):
""" PIPELINE for matching
:param {str: str} params:
"""
logging.info('running...')
# tl_expt.set_experiment_logger(params['path_expt'])
# tl_expt.create_subfolders(params['path_expt'], LIST_SUBDIRS)
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
if not os.path.isdir(params['path_output']):
os.mkdir(params['path_output'])
df_info = pd.read_csv(params['path_infofile'], index_col=0)
df_info = r_match.filter_table(df_info, params['path_images'])
df_info.dropna(inplace=True)
df_info = df_info[df_info['ellipse_Jaccard'] >= OVERLAP_THRESHOLD]
logging.info('filtered %i item in table' % len(df_info))
# execute over groups per stage
path_dir_imgs = os.path.dirname(params['path_images'])
for stage, df_stage in df_info.groupby('stage'):
perform_stage(df_stage, stage, path_dir_imgs, params['path_output'])
logging.info('DONE')
def parse_arg_params():
""" create simple arg parser with default values (input, results, dataset)
:return obj: argparse
"""
parser = argparse.ArgumentParser()
parser.add_argument('-imgs',
'--path_images',
type=str,
required=True,
help='path to dir with annot',
default=PATH_IMAGES)
parser.add_argument('--label',
type=int,
required=False,
nargs='+',
help='labels to be replaced',
default=[-1])
parser.add_argument('--nb_jobs',
type=int,
required=False,
help='number of jobs in parallel',
default=NB_THREADS)
args = vars(parser.parse_args())
p_dir = tl_data.update_path(os.path.dirname(args['path_images']))
assert os.path.isdir(p_dir), 'missing folder: %s' % args['path_images']
args['path_images'] = os.path.join(p_dir,
os.path.basename(args['path_images']))
logging.info(tl_expt.string_dict(args, desc='ARG PARAMETERS'))
return args
def main(params):
""" PIPELINE for matching
:param {str: str} params:
"""
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
df_info = pd.read_csv(params['path_infofile'], sep='\t', index_col=0)
df_info = filter_table(df_info, params['path_ellipses'])
logging.info('filtered %i item in table' % len(df_info))
path_csv = os.path.join(params['path_output'], NAME_CSV_RESULTS)
list_evals = []
# get the folder
path_dir_csv = os.path.dirname(params['path_ellipses'])
_wrapper_match = partial(select_optimal_ellipse, path_dir_csv=path_dir_csv)
iterate = tl_expt.WrapExecuteSequence(_wrapper_match,
df_info.iterrows(),
nb_jobs=params['nb_jobs'])
for i, dict_row in enumerate(iterate):
list_evals.append(dict_row)
# every hundreds iteration do export
if i % 100 == 0:
pd.DataFrame(list_evals).to_csv(path_csv)
df_ellipses = pd.DataFrame(list_evals)
df_ellipses.to_csv(path_csv)
def main(params):
""" PIPELINE for new detections
:param {str: ...} params:
"""
tl_expt.set_experiment_logger(params['path_expt'])
# tl_expt.create_subfolders(params['path_expt'], LIST_SUBDIRS)
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
path_csv = os.path.join(params['path_expt'], NAME_CSV_TRIPLES)
df_paths = run_detect.get_csv_triplets(params['path_list'], path_csv,
params['path_images'],
params['path_segms'],
params['path_centers'], FORCE_RELOAD)
df_eval = df_paths.copy(deep=True)
for stage in params['stages']:
df_eval = evaluate_detection_stage(df_eval, stage,
params['path_infofile'],
params['path_expt'],
params['nb_jobs'])
if len(df_eval) > 0 and 'image' in df_eval.columns:
df_eval.set_index('image', inplace=True)
df_eval.to_csv(os.path.join(params['path_expt'], NAME_CSV_TRIPLES_STAT))
gc.collect()
time.sleep(1)
if len(df_eval) > 0:
df_stat = df_eval.describe().transpose()
logging.info('STATISTIC: \n %s', repr(df_stat))
df_stat.to_csv(os.path.join(params['path_expt'], NAME_CSV_STATISTIC))
def main(params):
""" PIPELINE for rotation
:param {str: str} params:
"""
logging.info('running...')
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
list_imgs = sorted(
[p for p in glob.glob(params['path_images']) if os.path.isfile(p)])
logging.info('found images: %i' % len(list_imgs))
if not os.path.isdir(params['path_output']):
os.mkdir(params['path_output'])
wrapper_object = partial(perform_orientation_swap,
path_out=params['path_output'])
dir_name = os.path.dirname(params['path_images'])
iterate = tl_expt.WrapExecuteSequence(wrapper_object,
list_imgs,
nb_jobs=params['nb_jobs'],
desc=dir_name)
list(iterate)
logging.info('DONE')
def main(params):
""" PIPELINE for new detections
:param {str: str} paths:
"""
logging.info('running...')
params = run_train.prepare_experiment_folder(params, FOLDER_EXPERIMENT)
# run_train.check_pathes_patterns(paths)
tl_expt.set_experiment_logger(params['path_expt'])
logging.info('COMPUTER: \n%s', repr(os.uname()))
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
tl_expt.create_subfolders(params['path_expt'], LIST_SUBFOLDER)
path_csv = os.path.join(params['path_expt'], NAME_CSV_TRIPLES)
df_paths = get_csv_triplets(params['path_list'],
path_csv,
params['path_images'],
params['path_segms'],
force_reload=FORCE_RERUN)
dict_classif = seg_clf.load_classifier(params['path_classif'])
params_clf = dict_classif['params']
params_clf.update(params)
logging.info(tl_expt.string_dict(params, desc='UPDATED PARAMETERS'))
# perform on new images
df_stat = pd.DataFrame()
wrapper_detection = partial(load_compute_detect_centers,
params=params_clf,
path_classif=params['path_classif'],
path_output=params['path_expt'])
iterate = tl_expt.WrapExecuteSequence(wrapper_detection,
df_paths.iterrows(),
nb_jobs=params['nb_jobs'])
for dict_center in iterate:
df_stat = df_stat.append(dict_center, ignore_index=True)
df_stat.to_csv(os.path.join(params['path_expt'],
NAME_CSV_TRIPLES_TEMP))
df_stat.set_index(['image'], inplace=True)
df_stat.to_csv(os.path.join(params['path_expt'], NAME_CSV_TRIPLES))
logging.info('STATISTIC: \n %s', repr(df_stat.describe()))
logging.info('DONE')
def main_predict(path_classif,
path_pattern_imgs,
path_out,
name='SEGMENT___',
params_local=None):
""" given trained classifier segment new images
:param str path_classif:
:param str path_pattern_imgs:
:param str path_out:
:param str name:
"""
logging.getLogger().setLevel(logging.INFO)
logging.info('running PREDICTION...')
assert path_pattern_imgs is not None
dict_classif = seg_clf.load_classifier(path_classif)
classif = dict_classif['clf_pipeline']
params = dict_classif['params']
if params_local is not None:
params.update({
k: params_local[k]
for k in params_local
if k.startswith('path_') or k.startswith('gc_')
})
path_out, path_visu = prepare_output_dir(path_pattern_imgs,
path_out,
name,
visual=params.get(
'visual', False))
tl_expt.set_experiment_logger(path_out)
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
paths_img = sorted(glob.glob(path_pattern_imgs))
logging.info('found %i images on path "%s"', len(paths_img),
path_pattern_imgs)
logging.debug('run prediction...')
show_debug_imgs = params.get('visual', False)
_wrapper_segment = partial(try_segment_image,
params=params,
classif=classif,
path_out=path_out,
path_visu=path_visu,
show_debug_imgs=show_debug_imgs)
list_img_path = list(zip([None] * len(paths_img), paths_img))
iterate = tl_expt.WrapExecuteSequence(_wrapper_segment,
list_img_path,
nb_jobs=params['nb_jobs'],
desc='segmenting images')
for _ in iterate:
gc.collect()
time.sleep(1)
logging.info('prediction DONE')
def main(params):
""" the main body containgn two approches:
1) segment each image indecently
2) estimate model over whole image sequence and estimate
:param {str: ...} params:
:return {str: ...}:
"""
logging.getLogger().setLevel(logging.DEBUG)
logging.info('running...')
reload_dir_config = (os.path.isfile(params['path_config']) or FORCE_RELOAD)
params = tl_expt.create_experiment_folder(
params,
dir_name=NAME_EXPERIMENT,
stamp_unique=EACH_UNIQUE_EXPERIMENT,
skip_load=reload_dir_config)
tl_expt.set_experiment_logger(params['path_exp'])
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_BASE)
if params['visual']:
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_DEBUG)
assert os.path.isfile(params['path_train_list']), \
'missing %s' % params['path_train_list']
dict_segms_gmm, dict_segms_group = {}, {}
df_paths = pd.read_csv(params['path_train_list'], index_col=0)
paths_img = df_paths['path_image'].tolist()
def path_expt(n):
return os.path.join(params['path_exp'], n)
# Segment as single model per image
dict_segms_gmm = experiment_single_gmm(params, paths_img,
path_expt(FOLDER_SEGM_GMM),
path_expt(FOLDER_SEGM_GMM_VISU))
gc.collect()
time.sleep(1)
dict_segms_group = experiment_group_gmm(params, paths_img,
path_expt(FOLDER_SEGM_GROUP),
path_expt(FOLDER_SEGM_GROUP_VISU))
gc.collect()
time.sleep(1)
df_ars = compare_segms_metric_ars(dict_segms_gmm,
dict_segms_group,
suffix='_gmm-group')
df_ars.to_csv(path_expt(NAME_CSV_ARS_CORES))
logging.info(df_ars.describe())
logging.info('DONE')
return params
def parse_arg_params():
""" create simple arg parser with default values (input, results, dataset)
:return obj: argparse
"""
parser = argparse.ArgumentParser()
parser.add_argument('-imgs',
'--path_images',
type=str,
required=True,
help='path to images',
default=PATH_IMAGES)
parser.add_argument('-out',
'--path_output',
type=str,
required=True,
help='path to output dir',
default=PATH_OUTPUT)
parser.add_argument('--label_old',
type=int,
required=False,
nargs='+',
help='labels to be replaced',
default=[0])
parser.add_argument('--label_new',
type=int,
required=False,
nargs='+',
help='new labels after replacing',
default=[0])
parser.add_argument('--nb_jobs',
type=int,
required=False,
help='number of jobs in parallel',
default=NB_THREADS)
args = vars(parser.parse_args())
for k in ['path_images', 'path_output']:
p_dir = tl_data.update_path(os.path.dirname(args[k]))
assert os.path.isdir(p_dir), 'missing folder: %s' % args[k]
args[k] = os.path.join(p_dir, os.path.basename(args[k]))
assert len(args['label_old']) == len(args['label_new']), \
'length of old (%i) and new (%i) labels should be same' \
% (len(args['label_old']), len(args['label_new']))
logging.info(tl_expt.string_dict(args, desc='ARG PARAMETERS'))
return args
def parse_arg_params():
""" create simple arg parser with default values (input, results, dataset)
:return obj: argparse
"""
parser = argparse.ArgumentParser()
parser.add_argument('-imgs',
'--path_images',
type=str,
required=True,
help='path to dir with images',
default=PATH_INPUT)
parser.add_argument('-out',
'--path_out',
type=str,
required=True,
help='path to output dir',
default=PATH_OUTPUT)
parser.add_argument('-clrs',
'--path_colors',
type=str,
required=False,
help='json with colour-label dict',
default=None)
parser.add_argument('--nb_jobs',
type=int,
required=False,
help='number of jobs in parallel',
default=NB_THREADS)
args = vars(parser.parse_args())
for n in ['path_images', 'path_out']:
p_dir = tl_data.update_path(os.path.dirname(args[n]))
assert os.path.isdir(p_dir), 'missing: %s' % args[n]
args[n] = os.path.join(p_dir, os.path.basename(args[n]))
if args['path_colors'] is not None:
args['path_colors'] = tl_data.update_path(args['path_colors'])
logging.info(tl_expt.string_dict(args, desc='ARG PARAMETERS'))
return args
def main(dict_paths, export_visual=EXPORT_VUSIALISATION, nb_jobs=NB_THREADS):
""" evaluate all segmentations in experiment folder
:param {str: str} paths: path to all required directories
:param bool export_visual: export visualisations
:param int nb_jobs: number threads in parralel
"""
logging.info('running in %i jobs...', nb_jobs)
logging.info(tl_expt.string_dict(dict_paths, desc='PATHS'))
list_results = sorted(glob.glob(os.path.join(dict_paths['results'], '*')))
list_results = sorted([
p for p in list_results
if os.path.isdir(p) and '___' not in os.path.basename(p)
and os.path.basename(p) not in SKIP_DIRS
])
tl_expt.create_subfolders(
dict_paths['results'],
[NAME_DIR_VISUAL_1, NAME_DIR_VISUAL_2, NAME_DIR_VISUAL_3])
df_all = pd.DataFrame()
wrapper_eval = partial(evaluate_folder,
dict_paths=dict_paths,
export_visual=export_visual)
iterate = tl_expt.WrapExecuteSequence(wrapper_eval,
list_results,
nb_jobs=nb_jobs)
for dict_eval in iterate:
df_all = df_all.append(dict_eval, ignore_index=True)
df_all.set_index(['method'], inplace=True)
df_all.sort_index(inplace=True)
logging.info('STATISTIC: \n %s', repr(df_all))
df_all.to_csv(
os.path.join(dict_paths['results'], NAME_CSV_STAT % 'OVERALL'))
logging.info('Done :]')
def parse_arg_params():
""" create simple arg parser with default values (input, results, dataset)
:return obj: argparse
"""
parser = argparse.ArgumentParser()
parser.add_argument('-imgs',
'--path_images',
type=str,
required=True,
help='path to dir with images',
default=PATH_IMAGES)
parser.add_argument('-m',
'--method',
type=str,
required=False,
help='method for quantisation color/position',
default='color',
choices=['color', 'position'])
parser.add_argument('-thr',
'--px_threshold',
type=float,
required=False,
help='percentage of pixels of a color to be removed',
default=THRESHOLD_INVALID_PIXELS)
parser.add_argument('--nb_jobs',
type=int,
required=False,
help='number of jobs in parallel',
default=NB_THREADS)
args = vars(parser.parse_args())
p_dir = tl_data.update_path(os.path.dirname(args['path_images']))
assert os.path.isdir(p_dir), 'missing folder: %s' % args['path_images']
args['path_images'] = os.path.join(p_dir,
os.path.basename(args['path_images']))
logging.info(tl_expt.string_dict(args, desc='ARG PARAMETERS'))
return args
def main_train(params):
""" PIPELINE for training
0) load triplets or create triplets from path to images, annotations
1) load precomputed data or compute them now
2) train classifier with hyper-parameters
3) perform Leave-One-Out experiment
:param {str: any} params:
"""
logging.info('run TRAINING...')
params = prepare_experiment_folder(params, FOLDER_EXPERIMENT)
tl_expt.set_experiment_logger(params['path_expt'])
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
with open(os.path.join(params['path_expt'], NAME_JSON_PARAMS), 'w') as f:
json.dump(params, f)
tl_expt.create_subfolders(params['path_expt'], LIST_SUBDIRS)
df_paths, _ = load_df_paths(params)
path_dump_data = os.path.join(params['path_expt'], NAME_DUMP_TRAIN_DATA)
if not os.path.isfile(path_dump_data) or FORCE_RECOMP_DATA:
dict_imgs, dict_segms, dict_slics, dict_points, dict_centers, \
dict_features, dict_labels, feature_names = \
dataset_load_images_segms_compute_features(params, df_paths,
params['nb_jobs'])
assert len(dict_imgs) > 0, 'missing images'
save_dump_data(path_dump_data, dict_imgs, dict_segms, dict_slics,
dict_points, dict_centers, dict_features, dict_labels,
feature_names)
else:
dict_imgs, dict_segms, dict_slics, dict_points, dict_centers, dict_features, \
dict_labels, feature_names = load_dump_data(path_dump_data)
if is_drawing(params['path_expt']) and EXPORT_TRAINING_DATA:
export_dataset_visual(params['path_expt'], dict_imgs, dict_segms,
dict_slics, dict_points, dict_labels,
params['nb_jobs'])
# concentrate features, labels
features, labels, sizes = seg_clf.convert_set_features_labels_2_dataset(
dict_features,
dict_labels,
drop_labels=[-1],
balance_type=params['balance'])
# remove all bad values from features space
features[np.isnan(features)] = 0
features[np.isinf(features)] = -1
assert np.sum(sizes) == len(labels), \
'not equal sizes (%d) and labels (%i)' \
% (int(np.sum(sizes)), len(labels))
# feature norm & train classification
nb_holdout = int(np.ceil(len(sizes) * CROSS_VAL_LEAVE_OUT_SEARCH))
cv = seg_clf.CrossValidatePSetsOut(sizes, nb_holdout)
classif, params['path_classif'] = seg_clf.create_classif_train_export(
params['classif'],
features,
labels,
cross_val=cv,
params=params,
feature_names=feature_names,
nb_search_iter=params['nb_classif_search'],
pca_coef=params.get('pca_coef', None),
nb_jobs=params['nb_jobs'],
path_out=params['path_expt'])
nb_holdout = int(np.ceil(len(sizes) * CROSS_VAL_LEAVE_OUT_EVAL))
cv = seg_clf.CrossValidatePSetsOut(sizes, nb_holdout)
seg_clf.eval_classif_cross_val_scores(params['classif'],
classif,
features,
labels,
cross_val=cv,
path_out=params['path_expt'])
seg_clf.eval_classif_cross_val_roc(params['classif'],
classif,
features,
labels,
cross_val=cv,
path_out=params['path_expt'])
if RUN_LEAVE_ONE_OUT:
experiment_loo(classif, dict_imgs, dict_segms, dict_centers,
dict_slics, dict_points, dict_features, feature_names)
logging.info('DONE')
def main(params):
""" the main body containgn two approches:
1) segment each image indecently
2) estimate model over whole image sequence and estimate
:param {str: ...} params:
:return {str: ...}:
"""
logging.getLogger().setLevel(logging.DEBUG)
show_visual = params.get('visual', False)
reload_dir_config = (os.path.isfile(params['path_config']) or FORCE_RELOAD)
params = tl_expt.create_experiment_folder(params,
dir_name=NAME_EXPERIMENT,
stamp_unique=params.get(
'unique',
EACH_UNIQUE_EXPERIMENT),
skip_load=reload_dir_config)
tl_expt.set_experiment_logger(params['path_exp'])
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_BASE)
if show_visual:
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_DEBUG)
paths_img = load_path_images(params)
assert len(paths_img) > 0, 'missing images'
def _path_expt(n):
return os.path.join(params['path_exp'], n)
# Segment as single model per image
path_visu = _path_expt(FOLDER_SEGM_GMM_VISU) if show_visual else None
dict_segms_gmm = experiment_single_gmm(params,
paths_img,
_path_expt(FOLDER_SEGM_GMM),
path_visu,
show_debug_imgs=show_visual)
gc.collect()
time.sleep(1)
# Segment as model ober set of images
if params.get('run_groupGMM', False):
path_visu = _path_expt(FOLDER_SEGM_GROUP_VISU) if show_visual else None
dict_segms_group = experiment_group_gmm(params,
paths_img,
_path_expt(FOLDER_SEGM_GROUP),
path_visu,
show_debug_imgs=show_visual)
else:
write_skip_file(_path_expt(FOLDER_SEGM_GROUP))
# write_skip_file(_path_expt(FOLDER_SEGM_GROUP_VISU))
dict_segms_group = None
if dict_segms_group is not None:
df_ars = compare_segms_metric_ars(dict_segms_gmm,
dict_segms_group,
suffix='_gmm-group')
df_ars.to_csv(_path_expt(NAME_CSV_ARS_CORES))
logging.info(df_ars.describe())
return params
def main_train(params):
""" the main composed from following steps:
1) load already computed data (features and labels) or compute them now
2) visualise labeled superpixels aka annotation
3) load or train classifier with hyper-parameters search
4) perform Leave-One-Out and Leave-P-Out experiments on images
:param {str: ...} params:
:return{str: ...} :
"""
logging.getLogger().setLevel(logging.DEBUG)
logging.info('running TRAINING...')
reload_dir_config = (os.path.isfile(params['path_config']) or FORCE_RELOAD)
params = tl_expt.create_experiment_folder(params, dir_name=NAME_EXPERIMENT,
stamp_unique=EACH_UNIQUE_EXPERIMENT,
skip_load=reload_dir_config)
tl_expt.set_experiment_logger(params['path_exp'])
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_BASE)
if params['visual']:
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_DEBUG)
df_stat = pd.DataFrame()
path_dump = os.path.join(params['path_exp'], NAME_DUMP_TRAIN_DATA)
if os.path.isfile(path_dump) and not FORCE_RECOMP_DATA:
dict_imgs, dict_annot, dict_slics, dict_features, dict_labels, \
dict_label_hist, feature_names = load_dump_data(path_dump)
else:
dict_imgs, dict_annot, dict_slics, dict_features, dict_labels, \
dict_label_hist, feature_names = \
dataset_load_images_annot_compute_features(params)
save_dump_data(path_dump, dict_imgs, dict_annot, dict_slics,
dict_features, dict_labels, dict_label_hist,
feature_names)
assert len(dict_imgs) > 1, 'training require at least 2 images'
dict_annot_slic = {n: np.asarray(dict_labels[n])[dict_slics[n]]
for n in dict_annot}
df = eval_segment_with_annot(params, dict_annot, dict_annot_slic,
dict_label_hist, NAME_CSV_SEGM_STAT_SLIC_ANNOT,
params['nb_jobs'])
df_stat = df_stat.append(get_summary(df, 'SLIC-annot'), ignore_index=True)
path_csv_stat = os.path.join(params['path_exp'], NAME_CSV_SEGM_STAT_RESULTS)
df_stat.set_index(['name']).to_csv(path_csv_stat)
if params['gc_use_trans']:
params['label_transitions'] = \
seg_gc.count_label_transitions_connected_segments(dict_slics,
dict_labels)
logging.info('summary on edge-label transitions: \n %s',
repr(params['label_transitions']))
for name in dict_labels:
weights = np.max(dict_label_hist[name], axis=1)
dict_labels[name][weights < params['label_purity']] = -1
logging.info('prepare features...')
# concentrate features, labels
features, labels, sizes = seg_clf.convert_set_features_labels_2_dataset(
dict_features, dict_labels, balance_type=params['balance'], drop_labels=[-1])
# drop "do not care" label which are -1
features = np.nan_to_num(features)
nb_holdout = max(1, int(round(len(sizes) * CROSS_VAL_LEAVE_OUT_SEARCH)))
params, classif, path_classif = load_train_classifier(params, features,
labels, feature_names,
sizes, nb_holdout)
# test classif on images
df_paths = pd.read_csv(params['path_train_list'], index_col=0)
paths_img = df_paths['path_image'].tolist()
perform_predictions(params, paths_img, classif)
# LEAVE ONE OUT
if RUN_CROSS_VAL_LOO:
df_stat = experiment_loo(params, df_stat, dict_annot, paths_img,
path_classif, path_dump)
# LEAVE P OUT
if RUN_CROSS_VAL_LPO:
df_stat = experiment_lpo(params, df_stat, dict_annot, paths_img,
path_classif, path_dump, nb_holdout)
logging.info('training DONE')
return params
def main_train(params):
""" the main composed from following steps:
1) load already computed data (features and labels) or compute them now
2) visualise labeled superpixels aka annotation
3) load or train classifier with hyper-parameters search
4) perform Leave-One-Out and Leave-P-Out experiments on images
:param {str: ...} params:
:return {str: ...}:
"""
logging.getLogger().setLevel(logging.DEBUG)
logging.info('running TRAINING...')
show_visual = params.get('visual', False)
reload_dir_config = (os.path.isfile(params.get('path_config', ''))
or FORCE_RELOAD)
params = tl_expt.create_experiment_folder(params,
dir_name=NAME_EXPERIMENT,
stamp_unique=params.get(
'unique',
EACH_UNIQUE_EXPERIMENT),
skip_load=reload_dir_config)
tl_expt.set_experiment_logger(params['path_exp'])
logging.info(tl_expt.string_dict(params, desc='PARAMETERS'))
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_BASE)
if show_visual:
tl_expt.create_subfolders(params['path_exp'], LIST_FOLDERS_DEBUG)
df_stat = pd.DataFrame()
path_dump = os.path.join(params['path_exp'], NAME_DUMP_TRAIN_DATA)
if os.path.isfile(path_dump) and not FORCE_RECOMP_DATA:
dict_imgs, dict_annot, dict_slics, dict_features, dict_labels, \
dict_label_hist, feature_names = load_dump_data(path_dump)
else:
dict_imgs, dict_annot, dict_slics, dict_features, dict_labels, \
dict_label_hist, feature_names = \
dataset_load_images_annot_compute_features(params,
show_visual)
save_dump_data(path_dump, dict_imgs, dict_annot, dict_slics,
dict_features, dict_labels, dict_label_hist,
feature_names)
assert len(dict_imgs) > 1, 'training require at least 2 images'
dict_annot_slic = {
n: np.asarray(dict_labels[n])[dict_slics[n]]
for n in dict_annot
}
df = eval_segment_with_annot(params, dict_annot, dict_annot_slic,
dict_label_hist,
NAME_CSV_SEGM_STAT_SLIC_ANNOT,
params.get('drop_labels',
None), params['nb_jobs'])
df_stat = df_stat.append(get_summary(df, 'SLIC-annot'), ignore_index=True)
path_csv_stat = os.path.join(params['path_exp'],
NAME_CSV_SEGM_STAT_RESULTS)
df_stat.set_index(['name']).to_csv(path_csv_stat)
if params['gc_use_trans']:
params['label_transitions'] = \
seg_gc.count_label_transitions_connected_segments(dict_slics,
dict_labels)
logging.info('summary on edge-label transitions: \n %s',
repr(params['label_transitions']))
path_purity_visu = os.path.join(params['path_exp'], FOLDER_SLIC_ANNOT) \
if show_visual else None
dict_labels = filter_train_with_purity(dict_imgs,
dict_labels,
dict_label_hist,
params['label_purity'],
dict_slics,
drop_labels=params.get(
'drop_labels', None),
path_visu=path_purity_visu,
nb_jobs=params['nb_jobs'])
logging.info('prepare features...')
# concentrate features, labels
features, labels, sizes = seg_clf.convert_set_features_labels_2_dataset(
dict_features,
dict_labels,
balance_type=params['balance'],
drop_labels=[-1, np.nan] + params.get('drop_labels', []))
# drop "do not care" label which are -1
features = np.nan_to_num(features)
nb_holdout = params.get('cross_val', CROSS_VAL_LEAVE_OUT_SEARCH)
nb_holdout = max(1, int(round(len(sizes) * nb_holdout))) # minimum is 1
nb_holdout = min(nb_holdout, int(len(sizes) / 2)) # max is half of the set
params, classif, path_classif = load_train_classifier(
params, features, labels, feature_names, sizes, nb_holdout)
def _path_expt(n):
return os.path.join(params['path_exp'], n)
# test classif. on training images
df_paths = pd.read_csv(params['path_train_list'], index_col=0)
df_paths.reset_index(inplace=True)
paths_img = df_paths['path_image'].tolist()
if RUN_TRAIN_PREDICT:
perform_train_predictions(params,
paths_img,
classif,
show_debug_imgs=show_visual)
else:
write_skip_file(_path_expt(FOLDER_TRAIN))
gc.collect()
time.sleep(1)
# LEAVE P OUT
if params.get('run_LPO', True):
idx_paths_img = list(
zip(df_paths.index.tolist(), df_paths['path_image'].tolist()))
df_stat = experiment_lpo(params,
df_stat,
dict_annot,
idx_paths_img,
path_classif,
path_dump,
nb_holdout,
show_debug_imgs=show_visual)
else:
write_skip_file(_path_expt(FOLDER_LPO))
# write_skip_file(_path_expt(FOLDER_LPO_VISU))
logging.info('Statistic: \n %s', repr(df_stat.describe()))
logging.info('training DONE')
return params
