text
stringlengths 1
93.6k
|
|---|
current_step = diffusion.begin_step
|
current_epoch = diffusion.begin_epoch
|
n_epoch = opt['train']['n_epoch']
|
if opt['path']['resume_state']:
|
logger.info('Resuming training from epoch: {}, iter: {}.'.format(current_epoch, current_step))
|
diffusion.set_new_noise_schedule(opt['model']['beta_schedule'][opt['phase']], schedule_phase=opt['phase'])
|
val_dice = 0
|
if opt['phase'] == 'train':
|
while current_epoch < n_epoch:
|
current_epoch += 1
|
for istep, train_data in enumerate(train_loader):
|
iter_start_time = time.time()
|
current_step += 1
|
diffusion.feed_data(train_data)
|
diffusion.optimize_parameters()
|
# log
|
if (istep+1) % opt['train']['print_freq'] == 0:
|
logs = diffusion.get_current_log()
|
t = (time.time() - iter_start_time) / batchSize
|
visualizer.print_current_errors(current_epoch, istep+1, training_iters, logs, t, 'Train')
|
visualizer.plot_current_errors(current_epoch, (istep+1) / float(training_iters), logs)
|
visuals = diffusion.get_current_visuals()
|
visualizer.display_current_results(visuals, current_epoch, True)
|
# validation
|
if (current_step+1) % opt['train']['val_freq'] == 0:
|
diffusion.test(continous=False)
|
visuals = diffusion.get_current_visuals(isTrain=False)
|
visualizer.display_current_results(visuals, current_epoch, True)
|
if current_epoch % opt['train']['save_checkpoint_epoch'] == 0:
|
logger.info('Saving models and training states.')
|
diffusion.save_network(current_epoch, current_step)
|
dice_per_case_score = 0
|
for idata, val_data in enumerate(val_loader):
|
diffusion.feed_data(val_data)
|
diffusion.test_segment()
|
visuals = diffusion.get_current_segment()
|
predseg = visuals['test_V'].squeeze().numpy()
|
predseg = (predseg + 1) / 2.
|
predseg = (predseg > 0.5).astype(bool)
|
label = val_data['F'].cpu().squeeze().numpy()
|
label = (label + 1) / 2.
|
label = (label > 0.5).astype(bool)
|
dice = (visualizer.calculate_score(label, predseg, "dice"))
|
dice_per_case_score += dice
|
dice_case = (dice_per_case_score) / valid_iters
|
if dice_case >= val_dice:
|
val_dice = dice_case
|
diffusion.save_network(current_epoch, current_step, seg_save=True, dice=round(val_dice, 4))
|
# save model
|
logger.info('End of training.')
|
else:
|
logger.info('Begin Model Evaluation.')
|
result_path = '{}'.format(opt['path']['results'])
|
os.makedirs(result_path, exist_ok=True)
|
file = open(result_path + 'test_score.txt', 'w')
|
dice_per_case_score = []
|
prec_per_case_score = []
|
jacc_per_case_score = []
|
for idata, val_data in enumerate(val_loader):
|
dataInfo = val_data['P'][0].split('\\')[-1][:-4]
|
time1 = time.time()
|
diffusion.feed_data(val_data)
|
diffusion.test_segment()
|
time2 = time.time()
|
visuals = diffusion.get_current_segment()
|
predseg = visuals['test_V'].squeeze().numpy()
|
predseg = (predseg + 1) / 2.
|
predseg = (predseg > 0.5).astype(bool)
|
label = val_data['F'].cpu().squeeze().numpy()
|
label = (label + 1) / 2.
|
label = (label > 0.5).astype(bool)
|
data = val_data['A'].cpu().squeeze().numpy()
|
data = (data+1)/2.
|
savePath = os.path.join(result_path, '%d_data.png' % idata)
|
save_image(data * 255, savePath)
|
savePath = os.path.join(result_path, '%d_pred.png' % (idata))
|
save_image(predseg * 255, savePath)
|
savePath = os.path.join(result_path, '%d_label.png' % (idata))
|
save_image(label * 255, savePath)
|
dice = (visualizer.calculate_score(label, predseg, "dice"))
|
prec = (visualizer.calculate_score(label, predseg, "prec"))
|
jacc = (visualizer.calculate_score(label, predseg, "jacc"))
|
file.write('%04d: process image... %03s | Dice=%f | Prec=%f | Jacc=%f \n' % (idata, dataInfo, dice, prec, jacc))
|
print('%04d: process image... %s' % (idata, dataInfo))
|
dice_per_case_score.append(dice)
|
prec_per_case_score.append(prec)
|
jacc_per_case_score.append(jacc)
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.