Test

loader.py

@@ -42,6 +42,63 @@ def __init__(self):
  print("Training set size: ", len(self.training_dataset))
  print("Validation set size: ", len(self.validation_dataset))
 
+ def imgs_input_fn(filenames, labels=None, perform_shuffle=False, repeat_count=1, batch_size=1):
+ def _parse_function(filename, label):
+ seq = filename[0]
+ frame = filename[1]
+
+ camera1_path = os.path.join(DATASET_DIR, "{:02}".format(seq), "image_0", "{:06}.png".format(frame))
+ camera2_path = os.path.join(DATASET_DIR, "{:02}".format(seq), "image_1", "{:06}.png".format(frame))
+ camera1_path_next = os.path.join(DATASET_DIR, "{:02}".format(seq), "image_0", "{:06}.png".format(frame + 1))
+ camera2_path_next = os.path.join(DATASET_DIR, "{:02}".format(seq), "image_1", "{:06}.png".format(frame + 1))
+ def get_image(filename):
+ image_string = tf.read_file(filename)
+ image = tf.image.decode_image(image_string, channels=1)
+ image.set_shape([None, None, None])
+ image = tf.image.resize_images(image, [HEIGHT, WIDTH])
+ image = tf.subtract(image, 127.5) # Zero-center by mean pixel
+ image = tf.divide(image, 127.5)
+ image.set_shape([HEIGHT, WIDTH, 3])
+ return image
+
+ image1 = get_image(camera1_path)
+ image2 = get_image(camera2_path)
+ image3 = get_image(camera1_path_next)
+ image4 = get_image(camera2_path_next)
+
+
+ d = dict([image1, image2, image3, image4]), label
+ return d
+
+ if labels is None:
+ labels = [0]*len(filenames)
+ labels=np.array(labels)
+ # Expand the shape of "labels" if necessory
+ if len(labels.shape) == 1:
+ labels = np.expand_dims(labels, axis=1)
+ filenames = tf.constant(filenames)
+ labels = tf.constant(labels)
+ labels = tf.cast(labels, tf.float32)
+ dataset = tf.data.Dataset.from_tensor_slices((filenames, labels))
+ dataset = dataset.map(_parse_function)
+ if perform_shuffle:
+ # Randomizes input using a window of 256 elements (read into memory)
+ dataset = dataset.shuffle(buffer_size=256)
+ dataset = dataset.repeat(repeat_count) # Repeats dataset this # times
+ dataset = dataset.batch(batch_size) # Batch size to use
+ iterator = dataset.make_one_shot_iterator()
+ batch_features, batch_labels = iterator.get_next()
+ return batch_features, batch_labels
+
+ def get_dataset(self):
+ imgs = []
+ labels = []
+ for data in self.training_dataset:
+ imgs.append(data.get_loc())
+ labels.append(data.get_matrix())
+ return imgs_input_fn(imgs,labels)
+
+
  def visualize(self, dataset):
  plot_numbers = [[], [], [], [], [], []]
  for data in dataset:
@@ -183,6 +240,9 @@ def __init__(self, sequence_id, frame_id, matrix1, matrix2):
  v = self.rotationMatrixToEulerAngles(rotation[:3,:3])
  self.matrix = np.array([self.translation_mat[0], self.translation_mat[1], self.translation_mat[2], v[0], v[1], v[2]])
 
+ def get_loc(self):
+ return [self.sequence_id, self.frame_id]
+
  def get_matrix(self):
  return self.matrix
 

model.py

@@ -43,7 +43,7 @@ def train(self, dataset, epochs, iterations, batch_size):
  with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
  opt = optimizer.minimize(loss)
 
- saver = tf.train.Saver(max_to_keep=10000)
+ saver = tf.train.Saver(tf.global_variables(), max_to_keep=10000)
 
  idx = 0
  while(os.path.exists(os.path.join(LOG_DIR, "egomotion" +str(idx)))):
@@ -63,6 +63,9 @@ def train(self, dataset, epochs, iterations, batch_size):
  sess = tf.Session(config=config)
  sess.run(tf.global_variables_initializer())
 
+ batch = dataset.get_dataset()
+ first_batch = sess.run(batch)
+ print(first_batch)
  for epoch in range(epochs):
 
  for iter in range(iterations):