# Calculate and log other batch metrics
cd_corrects = (
100 *
(cd_preds.squeeze().byte() == labels.squeeze().byte()).sum() /
(labels.size()[0] * (opt.patch_size**2)))
cd_train_report = prfs(labels.data.cpu().numpy().flatten(),
cd_preds.data.cpu().numpy().flatten(),
average='binary',
pos_label=1)
train_metrics = set_metrics(train_metrics, cd_loss, cd_corrects,
cd_train_report, scheduler.get_lr())
# log the batch mean metrics
mean_train_metrics = get_mean_metrics(train_metrics)
for k, v in mean_train_metrics.items():
writer.add_scalars(str(k), {'train': v}, total_step)
# clear batch variables from memory
del batch_img1, batch_img2, labels
scheduler.step()
logging.info("EPOCH {} TRAIN METRICS".format(epoch) +
str(mean_train_metrics))
"""
Begin Validation
"""
model.eval()
with torch.no_grad():
batch_img1 = autograd.Variable(batch_img1).float().to(dev)
batch_img2 = autograd.Variable(batch_img2).float().to(dev)
labels = autograd.Variable(labels).long().to(dev)
# Get predictions and calculate loss
cd_preds = model(batch_img1, batch_img2)
cd_preds = cd_preds[-1]
_, cd_preds = torch.max(cd_preds, 1)
# Calculate and log other batch metrics
cd_corrects = (
100 *
(cd_preds.squeeze().byte() == labels.squeeze().byte()).sum() /
(labels.size()[0] * (opt.patch_size**2)))
cd_val_report = prfs(labels.data.cpu().numpy().flatten(),
cd_preds.data.cpu().numpy().flatten(),
average='binary',
pos_label=1)
test_metrics = set_test_metrics(val_metrics, cd_corrects,
cd_val_report)
# log the batch mean metrics
mean_test_metrics = get_mean_metrics(test_metrics)
# clear batch variables from memory
del batch_img1, batch_img2, labels
print("EPOCH VALIDATION METRICS", mean_test_metrics)