def start(self):
config , email_json = ConfigLoader.get_config()
#patch_pred = PatchPredictions(config)
if config['ensemble']:
pred = Ensemble_Predictions(config)
elif config['model_name'] in ['unet']:
pred = PatchPredictions(config)
else:
pred = PixelPredictions(config)
if config['change_targets']:
ind_dirs = PathUtils.get_indicator_directories(None)
target_ids = list(range(-1,len(ind_dirs)))
# target_ids = list(range(0,3))
scores = {}
for id in target_ids:
print(f'target id {id}')
pred.set_target_paths(id)
max_score, avg_score = pred.train_predict()
scores[id] = {"max_score":round(max_score,5), "avg_score":round(avg_score,5)}
for key,value in scores.items():
print(f"For target id {key} the max score is {value['max_score']} and avg score is {value['avg_score']}")
else:
# train_gen, test_gen = pred.get_data_generators()
pred.train_predict()
def __init__(self, output_dir, indicators_path, patch_shape,
img_downscale_factor, tuning):
self.output_dir = output_dir
self.patch_shape = patch_shape, patch_shape
self.indicators_path = indicators_path
self.indicator_directories = PathUtils.get_indicator_directories(
self.indicators_path)
self.img_downscale_factor = img_downscale_factor
self.tuning = tuning
def __init__(self,
config,
model_name=None,
output_dir=None,
target_id=None):
super().__init__(config, model_name, output_dir)
self.patches_path, self.patch_img_ref_path, self.indicators_path, img_ref_csv, self.ref_data_path = PathUtils.get_paths_for_patches(
self.config)
self.indicator_directories = PathUtils.get_indicator_directories(
self.indicators_path)
self.set_target_paths(target_id or config['target_id'])
self.starting_index, self.ending_index = JsonLoader.get_data_size(
self.config)
self._prepare_img_refs(self.patch_img_ref_path)
def start(self):
img_ref_csv_path, ref_data_path, targets_path, indicators_path = PathUtils.get_paths(
self.config)
irb = ImgRefBuilder(img_ref_csv_path)
starting_index, ending_index = JsonLoader.get_data_size(self.config)
img_refs = irb.get_img_ref(starting_index, ending_index)
patches_folder = self.config['path']['outputs'] + "patches/"
output_dir = FolderUtils.create_patch_output_folder(
self.patch_shape, self.img_downscale_factor, patches_folder,
PathUtils.get_indicator_directories(indicators_path),
self.config['tuning'], self.config['data_year'],
self.config['data_prefix'])
LogUtils.init_log(output_dir)
pg = PatchGenerator(output_dir=output_dir,
indicators_path=indicators_path,
img_downscale_factor=self.img_downscale_factor,
patch_shape=self.patch_shape,
tuning=self.config["tuning"])
pg.create_img_patches(img_refs, targets_path)
def __init__(self, config, model_name=None, output_dir=None):
super().__init__(config, model_name, output_dir)
img_ref_csv_path, self.ref_data_path, self.targets_path, self.indicators_path = PathUtils.get_paths(self.config)
self.indicator_directories = PathUtils.get_indicator_directories(self.indicators_path)
self.image_downscale_factor = config['image_downscale_factor']
self._prepare_img_refs(img_ref_csv_path)