def main(argv):
parser = argparse.ArgumentParser(
description='Display image readable with tensorflow.')
parser.add_argument('filename', type=str, help='Image file to display.')
args = parser.parse_args()
if ctfi.is_image(args.filename) == False:
sys.exit(-1)
image = ctfi.load(args.filename, channels=3)
image = tf.expand_dims(image, 0)
dx, dy = tf.image.image_gradients(image)
dxr, dxg, dxb = tf.split(dx, 3, 3)
dyr, dyg, dyb = tf.split(dy, 3, 3)
strides = [1, 1, 1, 1]
padding = "SAME"
#reconstructed = tf.nn.conv2d_transpose(dxr + dyr, tf.ones([3,3,1,1], dtype=tf.float32),[1,32,32,1],strides,padding)# + tf.nn.conv2d(dy, tf.ones([3,3,1,3], dtype=tf.float32),strides,padding)
#reconstructed = tf.concat([tf.nn.conv2d_transpose(c, tf.ones([1,32,1,1], dtype=tf.float32),[1,32,32,1],strides,padding) for c in tf.split(dx,3,3)],3)
#reconstructed += tf.concat([tf.nn.conv2d_transpose(c, tf.ones([32,1,1,1], dtype=tf.float32),[1,32,32,1],strides,padding) for c in tf.split(dy,3,3)],3)
fig, ax = plt.subplots(2, 2)
ax[0, 0].imshow(image[0].numpy())
ax[0, 1].imshow(dx[0] + dy[0].numpy())
ax[1, 0].imshow(dx[0].numpy())
ax[1, 1].imshow(dy[0].numpy())
plt.show()
def main(argv):
filename = os.path.join(git_root, 'data', 'images', 'tile_8_14.jpeg')
if ctfi.is_image(filename):
image = ctfi.load(filename, width=1024, height=1024, channels=3)
else:
image = np.random.rand(1024, 1024, 3)
# Using eager execution
fig, ax = plt.subplots()
plt.imshow(image.numpy())
plt.show()
def main(argv):
parser = argparse.ArgumentParser(
description='Compute latent code for image patch by model inference.')
parser.add_argument('export_dir',
type=str,
help='Path to saved model to use for inference.')
parser.add_argument('filename',
type=str,
help='Image file or numpy array to run inference on.')
parser.add_argument('--output',
type=str,
help='Where to store the output.')
args = parser.parse_args()
predict_fn = predictor.from_saved_model(args.export_dir)
# Extract patch size and latent space size from the model identifier
patch_size = ctfsm.determine_patch_size(args.export_dir)
latent_space_size = ctfsm.determine_latent_space_size(args.export_dir)
image = None
# Check if it is image or numpy array data
if ctfi.is_image(args.filename):
image = ctfi.load(args.filename).numpy()
elif cutil.is_numpy_format(args.filename):
image = np.load(args.filename)
else:
sys.exit(3)
# Resize image to match size required by the model
image = np.resize(image, [patch_size, patch_size, 3])
batch = np.expand_dims(image, 0)
# Make predictions
pred = predict_fn({
'fixed': batch,
'moving': np.random.rand(1, patch_size, patch_size, 3),
'embedding': np.random.rand(1, 1, 1, latent_space_size)
})
latent_code = pred['latent_code_fixed']
print(latent_code)
if args.output:
with open(args.output, 'w') as f:
json.dump(
{
'filename': args.filename,
'model': args.export_dir,
'latent_code': latent_code.tolist()
}, f)
def main(argv):
parser = argparse.ArgumentParser(description='Display image readable with tensorflow.')
parser.add_argument('filename',type=str,help='Image file to display.')
args = parser.parse_args()
if ctfi.is_image(args.filename) == False:
sys.exit(-1)
image = ctfi.load(args.filename, channels=3)
fig, ax = plt.subplots()
plt.imshow(image.numpy())
plt.show()