def fold_evaluate_loop(valid_data_split, fold_id):
valid_pipe_input = {
'input': {
'meta': valid_data_split
},
'callback_input': {
'meta_valid': None
}
}
valid_ids = valid_data_split[ID_COLUMN].tolist()
LOGGER.info('Start pipeline transform on valid')
pipeline_network = unet(config=CONFIG,
suffix='_fold_{}'.format(fold_id),
train_mode=False)
pipeline_postprocessing = pipelines.mask_postprocessing(
config=CONFIG, suffix='_fold_{}'.format(fold_id))
pipeline_network.clean_cache()
pipeline_postprocessing.clean_cache()
predicted_masks_valid = pipeline_network.transform(valid_pipe_input)
valid_pipe_masks = {'input_masks': predicted_masks_valid}
output_valid = pipeline_postprocessing.transform(valid_pipe_masks)
utils.clean_object_from_memory(pipeline_network)
y_pred_valid = output_valid['binarized_images']
y_true_valid = utils.read_masks(valid_data_split[Y_COLUMN].values)
iou, iout = calculate_scores(y_true_valid, y_pred_valid)
predicted_masks_valid = predicted_masks_valid['resized_images']
return iou, iout, (valid_ids, predicted_masks_valid)
def fold_fit_evaluate_loop(train_data_split, valid_data_split, fold_id):
train_pipe_input = {'input': {'meta': train_data_split
},
'callback_input': {'meta_valid': valid_data_split
}
}
valid_pipe_input = {'input': {'meta': valid_data_split
},
'callback_input': {'meta_valid': None
}
}
valid_ids = valid_data_split[ID_COLUMN].tolist()
LOGGER.info('Start pipeline fit and transform on train')
config = add_fold_id_suffix(CONFIG, fold_id)
pipeline_network = network(config=config, suffix='_fold_{}'.format(fold_id), train_mode=True)
pipeline_network.clean_cache()
pipeline_network.fit_transform(train_pipe_input)
utils.clean_object_from_memory(pipeline_network)
LOGGER.info('Start pipeline transform on valid')
pipeline_network = network(config=config, suffix='_fold_{}'.format(fold_id), train_mode=False)
pipeline_network.clean_cache()
predicted_masks_valid = pipeline_network.transform(valid_pipe_input)
utils.clean_object_from_memory(pipeline_network)
y_pred_valid = predicted_masks_valid['resized_images']
y_true_valid = valid_data_split[Y_COLUMN].values
auc = calculate_scores(y_true_valid, y_pred_valid)
return auc, (valid_ids, y_pred_valid)
def fold_evaluate_predict_loop(valid_data_split, test, fold_id):
iou, iout, predicted_masks_valid = fold_evaluate_loop(valid_data_split, fold_id)
test_pipe_input = {'input': {'meta': test
},
'callback_input': {'meta_valid': None
}
}
pipeline_network = network(config=CONFIG, suffix='_fold_{}'.format(fold_id), train_mode=False)
LOGGER.info('Start pipeline transform on test')
pipeline_network.clean_cache()
predicted_masks_test = pipeline_network.transform(test_pipe_input)
utils.clean_object_from_memory(pipeline_network)
predicted_masks_test = predicted_masks_test['resized_images']
return iou, iout, predicted_masks_valid, predicted_masks_test
