| |
| from time import time, sleep |
| from datetime import datetime |
| import os |
| import sys |
| import pandas as pd |
| import numpy as np |
| import matplotlib.pyplot as plt |
|
|
|
|
| class Logger(): |
| def __init__(self, output_folder): |
| self.output_folder = output_folder |
| os.makedirs(self.output_folder, exist_ok=True) |
| |
| timestamp = datetime.now() |
| self.log_file = os.path.join( |
| self.output_folder, |
| "training_log_%d_%d_%d_%02.0d_%02.0d_%02.0d.txt" % |
| (timestamp.year, timestamp.month, timestamp.day, timestamp.hour, timestamp.minute, |
| timestamp.second) |
| ) |
| with open(self.log_file, 'w') as f: |
| f.write("Starting... \n") |
| |
| def print_to_log_file(self, *args, also_print_to_console=True, add_timestamp=True): |
| timestamp = time() |
| dt_object = datetime.fromtimestamp(timestamp) |
|
|
| if add_timestamp: |
| args = ("%s:" % dt_object, *args) |
| |
| successful = False |
| max_attempts = 5 |
| ctr = 0 |
| while not successful and ctr < max_attempts: |
| try: |
| with open(self.log_file, 'a+') as f: |
| for a in args: |
| f.write(str(a)) |
| f.write(" ") |
| f.write("\n") |
| successful = True |
| except IOError: |
| print("%s: failed to log: " % datetime.fromtimestamp(timestamp), sys.exc_info()) |
| sleep(0.5) |
| ctr += 1 |
| if also_print_to_console: |
| print(*args) |
| |
| |
| def poly_lr(epoch, max_epochs, initial_lr, exponent=0.9): |
| return initial_lr * (1 - epoch / max_epochs)**exponent |
|
|
|
|
| def split_prostatedataset(data_root, patientid_path, center=1, seed=0, random_validation=False): |
| all_centers = ['A-ISBI', 'B-ISBI_1.5', 'C-I2CVB', 'D-UCL', 'E-BIDMC', 'F-HK'] |
| train_centers = [all_centers[center-1]] |
| patientid = pd.read_csv(patientid_path) |
| all_patientid = [] |
| train_patientid = [] |
| val_patientid = [] |
| for c in train_centers: |
| all_patientid.extend(patientid[patientid.center==c]['patientid'].values.tolist()) |
|
|
| if random_validation: |
| np.random.seed(seed) |
| np.random.shuffle(all_patientid) |
| train_patientid = sorted(all_patientid[:int(len(all_patientid)*0.9)]) |
| val_patientid = sorted(all_patientid[int(len(all_patientid)*0.9):]) |
| else: |
| train_patientid = all_patientid[:int(len(all_patientid)*0.9)] |
| val_patientid = all_patientid[int(len(all_patientid)*0.9):] |
|
|
| train_patientid = [x for x in os.listdir(data_root) if int(x[9:12]) in train_patientid] |
| val_patientid = [x for x in os.listdir(data_root) if int(x[9:12]) in val_patientid] |
|
|
| |
| ood_centers = list(set(all_centers) - set(train_centers)) |
| test_patientid = [] |
| for c in ood_centers: |
| test_patientid.extend(patientid[patientid.center==c]['patientid'].values.tolist()) |
| |
| return train_patientid, val_patientid, test_patientid |
|
|
|
|
| def plot_loss(train_losses, val_losses, model_save_path): |
| |
| plt.plot(train_losses) |
| plt.plot(val_losses) |
| plt.title('Dice CE Loss') |
| plt.xlabel('Epoch') |
| plt.ylabel('Loss') |
| plt.savefig(os.path.join(model_save_path, 'train_loss.png')) |
| plt.close() |
|
|
|
|
