image = read_image(imgs[i], normalise=True)
image = resize_image(centre_crop(image), args.size)
# Compute the saliency map
grad = localiser.saliency(image)
grad = median(grad, disk(3))
if args.scale:
grad = np.clip(1.0 * grad / args.scale, 0, 1)
else:
grad = 1.0 * grad / np.max(grad)
if args.red:
grad = np.stack([
np.ones(grad.shape),
np.zeros(grad.shape),
np.zeros(grad.shape), grad
],
axis=2)
if args.output:
fname = os.path.join(args.output, os.path.basename(imgs[i]))
scipy.misc.toimage(grad, cmin=0, cmax=1).save(fname)
progress_bar(1.0 * (i + 1) / n_images, 30, text='Computing')
else:
# Display image
plt.imshow(grad, cmap='gray')
plt.show()
if n_images > 1:
print('')
[train_loss]+train_summaries, feed_dict={x: train_images_feed, y: train_labels_feed})
for res in results[1:]:
summary_writer.add_summary(res, i)
if args.verbose:
print('{} (training): Loss = {:.6f}'.format(i, results[0]))
if args.validate and (i % n_disp_validation == 0):
val_images_feed, val_labels_feed = validation_reader.next_batch()
results = sess.run(
[validation_loss]+validation_summaries, feed_dict={x: val_images_feed, y: val_labels_feed})
for res in results[1:]:
summary_writer.add_summary(res, i)
if args.verbose:
print('{} (validation): Loss = {:.6f}'.format(i, results[0]))
if not args.verbose:
progress_bar(1.0*(i+1-args.iter_start)/args.n_iters, 30, text='Training',
epilog='iter {}'.format(i))
if i+1 in args.save_iters:
save_path = os.path.join(args.save_dir, args.name + '_iter{}.ckpt'.format(i+1))
saver.save(sess, save_path)
print('')
# Save the model
save_path = os.path.join(args.save_dir, args.name + '.ckpt')
saver.save(sess, save_path)
print("Final model saved in file: %s" % save_path)
with Localiser(args.model, uncertainty=args.uncertainty) as localiser:
for i in range(n_images):
images_feed, labels_feed = test_reader.next_batch()
# Make prediction
predicted = localiser.localise(images_feed)
positions = np.concatenate(
(positions, np.asarray([predicted['x']])))
orientations = np.concatenate(
(orientations, np.asarray([predicted['q']])))
if args.uncertainty:
std_x.append(predicted['std_x'])
std_q.append(predicted['std_q'])
progress_bar(1.0 * (i + 1) / n_images, 30, text='Localising')
print('')
else:
positions = positions_gt
orientations = orientations_gt
def draw_segment(start, end, color='black', lw=1):
plt.plot([start[0], end[0]], [start[1], end[1]], [start[2], end[2]],
linestyle='-',
color=color,
lw=lw)
def draw_sphere(pos, r, color='black', lw=1):
u = np.linspace(0, 2 * np.pi, 100)
