global_step=global_step)
print('save a checkpoint at ' + opt.checkpoint_path + 'model-' +
str(it))
print('start testing {} samples...'.format(test_samples))
for ti in range(test_samples):
x_batch, y_batch = next(test_generator)
# tensorflow wants a different tensor order
feed_dict = {img: x_batch, label: y_batch}
loss, pred_logits = sess.run([log_loss, pred],
feed_dict=feed_dict)
pred_map_batch = np.argmax(pred_logits, axis=3)
# import pdb; pdb.set_trace()
for pred_map, y in zip(pred_map_batch, y_batch):
mean_iou, pixel_acc, dice = vis.add_sample(
pred_map, y_batch[0])
vis.compute_scores(suffix=it)
x_batch, y_batch = next(train_generator)
# w_map = define_map(y_batch[0])
# exclude whole-backgroud images
feed_dict = {
img: x_batch,
label: y_batch
# weight_map: w_map
}
_, loss, summary, lr, pred_logits = sess.run(
[train_step, log_loss, summary_merged, learning_rate, pred],
feed_dict=feed_dict)
global_step.assign(it).eval()
train_writer.add_summary(summary, it)
tf.train.latest_checkpoint(opt.load_from_checkpoint))
print('--> load from checkpoint ' + opt.load_from_checkpoint)
except Exception as ex:
print('Unable to load checkpoint ...', ex)
sys.exit(0)
dice_score = 0
for it in range(0, test_samples):
x_batch, y_batch = next(test_generator)
# tensorflow wants a different tensor order
feed_dict = {img: x_batch, label: y_batch}
loss, pred_logits = sess.run([cross_entropy_loss, pred],
feed_dict=feed_dict)
pred_map = np.argmax(pred_logits[0], axis=2)
score = vis.add_sample(pred_map, y_batch[0])
im, gt = deprocess(x_batch[0], dataset_mean, dataset_std, y_batch[0])
# im = Image.fromarray(im)
# im.save(os.path.join(opt.load_from_checkpoint, '{0:}.png'.format(it)))
# gt = Image.fromarray(gt, 'L')
# gt.save(os.path.join(opt.load_from_checkpoint, '{0:}_{1:.3f}.png'.format(it, score)))
vis.save_seg(pred_map,
name='{0:}_{1:.3f}.png'.format(it, score),
im=im,
gt=gt)
print('[iter %f]: loss=%f, meanIU=%f' % (it, loss, score))
vis.compute_scores()