l3_finder_tester.py

import tensorflow as tf
from keras.backend.tensorflow_backend import set_session

from ct_slice_detection.io.dataloader import DataLoader
from ct_slice_detection.io.parameters import parse_inputs
from ct_slice_detection.models import Models
from ct_slice_detection.utils.testing_utils import *


def main():

    args = parse_inputs()

    print(args)

    # GPU allocation options
    config = tf.ConfigProto()
    config.gpu_options.allow_growth = True
    config.gpu_options.visible_device_list = args.cuda_devices
    set_session(tf.Session(config=config))



    baseModel = Models(args.model_name)

    if args.do_crossval:
        test_data = DataLoader(args)
        test_data.load_train_val_split(do_cross_val=True)
        cross_val_data = test_data.cross_val_data.item()
        print(cross_val_data.keys())
        num_splits = args.n_splits
        for i in range(num_splits):
            train_index = cross_val_data[i]['train']
            val_index = cross_val_data[i]['val']
            test_data.split_data(train_index, val_index)

            model_name = args.model_name + '_cv_' + str(i+1) + '_of_' + str(num_splits)
            modelwrapper = baseModel(name=model_name,
                                     config=args,
                                     input_shape=args.model_input_shape,
                                     data_loader=test_data
                                     )

            modelwrapper.model.compile(optimizer='adam', loss='mse', metrics=['mae'])
            modelwrapper.load()

            with open(os.path.join(args.model_path,"time_log.txt"), "a") as file:
                start_time = time.time()
                predict_and_evaluate(args, test_data, modelwrapper, suffix='_cv_'+str(i))
                end_time = time.time()
                dt = end_time - start_time
                n = len(test_data.x_val)
                file.write("{}, total time for {}: {} , (avg: {}) \n ".format(model_name, n, dt, dt / n))
    else:
        test_data = DataLoader(args)

        if args.test_mode == 'eval':
            test_data.load_train_val_split()
            test_data.split_data(test_data.train_idx, test_data.val_idx)
        else:
            test_data.load_data()

        # Setup model
        modelwrapper = baseModel(name=args.model_name,
                                   config=args,
                                   input_shape=args.model_input_shape,
                                   data_loader=test_data
                                   )

        modelwrapper.model.compile(optimizer='adam', loss='mse', metrics=['mae'])
        modelwrapper.load()

        predict_and_evaluate(args, test_data, modelwrapper)




if __name__ == '__main__':
    main()