import os.path as osp import logging import shutil import time import argparse import os import utils import utils.options as option import utils.util as util from trainer.ExtensibleTrainer import ExtensibleTrainer from data import create_dataset, create_dataloader from tqdm import tqdm import torch import torchvision if __name__ == "__main__": #### options torch.backends.cudnn.benchmark = True want_metrics = False parser = argparse.ArgumentParser() parser.add_argument('-opt', type=str, help='Path to options YAML file.', default='../options/test_noisy_audio_clips_classifier.yml') opt = option.parse(parser.parse_args().opt, is_train=False) opt = option.dict_to_nonedict(opt) utils.util.loaded_options = opt util.mkdirs( (path for key, path in opt['path'].items() if not key == 'experiments_root' and 'pretrain_model' not in key and 'resume' not in key)) util.setup_logger('base', opt['path']['log'], 'test_' + opt['name'], level=logging.INFO, screen=True, tofile=True) logger = logging.getLogger('base') logger.info(option.dict2str(opt)) #### Create test dataset and dataloader test_loaders = [] for phase, dataset_opt in sorted(opt['datasets'].items()): if 'test' in phase: test_set = create_dataset(dataset_opt) if hasattr(test_set, 'wrapped_dataset'): test_set = test_set.wrapped_dataset test_loader = create_dataloader(test_set, dataset_opt, opt) logger.info('Number of test images: {:d}'.format(len(test_set))) test_loaders.append((dataset_opt['name'], test_loader)) model = ExtensibleTrainer(opt) # Remove all losses, since often labels will not be provided and this is not implied in test. for s in model.steps: s.losses = {} output_key = opt['eval']['classifier_logits_key'] labels = opt['eval']['output_labels'] path_key = opt['eval']['path_key'] output_base_dir = util.opt_get(opt, ['eval', 'output_dir'], None) output_file = open('classify_into_folders.tsv', 'a') step = 0 for test_set_name, test_loader in test_loaders: test_start_time = time.time() dataset_dir = osp.join(opt['path']['results_root'], opt['name']) util.mkdir(dataset_dir) for data in tqdm(test_loader): with torch.no_grad(): model.feed_data(data, 0) model.test() lbls = torch.nn.functional.softmax(model.eval_state[output_key][0].cpu(), dim=-1) for k, lbl in enumerate(lbls): lbl = labels[torch.argmax(lbl, dim=0)] src_path = data[path_key][k] output_file.write(f'{src_path}\t{lbl}\n') if output_base_dir is not None: dest = os.path.join(output_base_dir, lbl) os.makedirs(dest, exist_ok=True) shutil.copy(str(src_path), os.path.join(dest, f'{step}_{os.path.basename(str(src_path))}')) step += 1 output_file.flush() output_file.close()