def main(_):
current_time = datetime.now().strftime("%Y%m%d-%H%M")
outdir = os.path.join(
FLAGS.outdir, FLAGS.arch + '-' + current_time + '-' + str(os.getpid()))
trained_model = os.path.join(outdir, 'trained_model')
summary_dir = os.path.join(outdir, 'summary')
utils.make_dirs(trained_model)
utils.print_arguments(args=FLAGS.__flags,
log_fn=os.path.join(outdir, 'args.log'))
logging.basicConfig(format='%(asctime)s [%(levelname)s] %(message)s',
filename='{}/train.log'.format(outdir),
filemode='w',
level=logging.INFO)
train(trained_model, summary_dir)
# Process arguments
args = parser.parse_args()
if args.trim <= 0:
args.trim = None
if args.force_source_file == '':
args.force_source_file = None
if args.force_source_speaker == '':
args.force_source_speaker = None
if args.force_target_speaker == '':
args.force_target_speaker = None
if args.fn_list == '':
args.fn_list = 'list_seed' + str(
args.seed_input) + '_' + args.split + '.tsv'
# Print arguments
utils.print_arguments(args)
# Seed
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if args.device == 'cuda':
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
torch.cuda.manual_seed(args.seed)
########################################################################################################################
# Load model, pars, & check
print('Load stuff')
from models import blow
import argparse
import functools
import paddle
from utils.utils import add_arguments, print_arguments, get_lfw_list
from utils.utils import get_features, get_feature_dict, test_performance
parser = argparse.ArgumentParser(description=__doc__)
add_arg = functools.partial(add_arguments, argparser=parser)
add_arg('batch_size', int, 64, '训练的批量大小')
add_arg('model_path', str, 'models/mobilefacenet/infer/model',
'MobileFaceNet预测模型的路径')
add_arg('test_list_path', str, 'dataset/lfw_test.txt', '测试数据的数据列表路径')
args = parser.parse_args()
print_arguments(args)
# 加载模型
model = paddle.jit.load(args.model_path)
model.eval()
# 获取测试数据
img_paths = get_lfw_list(args.test_list_path)
features = get_features(model, img_paths, batch_size=args.batch_size)
fe_dict = get_feature_dict(img_paths, features)
accuracy, threshold = test_performance(fe_dict, args.test_list_path)
print('准确率为：%f, 最优阈值为：%f' % (accuracy, threshold))