2martens/masterthesis
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Simplified data set creation by removing not required labels

Browse Source 
Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens
2019-04-28 19:06:44 +02:00
parent 7a89a3cf7b
commit 9b508477f3
2 changed files with 14 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
src/twomartens/masterthesis/data.py
View File

@ -268,20 +268,15 @@ def load_scenenet_val(photo_paths: Sequence[Sequence[str]],
bbox[3]
])
if not labels:
continue
final_image_paths.append(image_path)
final_labels.extend(labels)
final_labels.append(labels)
length_dataset = len(final_image_paths)
labels_np = np.array(final_labels)
print(labels_np.shape)
path_dataset = tf.data.Dataset.from_tensor_slices(final_image_paths)
label_dataset = tf.data.Dataset.from_tensor_slices(final_labels)
dataset = tf.data.Dataset.zip((path_dataset, label_dataset))
dataset = dataset.apply(tf.data.experimental.shuffle_and_repeat(buffer_size=length_dataset, count=num_epochs))
# label_dataset = tf.data.Dataset.from_tensor_slices(final_labels)
# dataset = tf.data.Dataset.zip((path_dataset, label_dataset))
dataset = path_dataset.apply(tf.data.experimental.shuffle_and_repeat(buffer_size=length_dataset, count=num_epochs))
dataset = dataset.batch(batch_size=batch_size)
dataset = dataset.map(_load_images_ssd_callback(resized_shape))
@ -289,8 +284,8 @@ def load_scenenet_val(photo_paths: Sequence[Sequence[str]],
def _load_images_ssd_callback(resized_shape: Sequence[int]) \
-> Callable[[Sequence[str], Sequence[Sequence[Tuple[int, int, int, int, int]]]],
Tuple[tf.Tensor, Sequence[Sequence[Tuple[int, int, int, int, int]]]]]:
-> Callable[[Sequence[str]],
Tuple[tf.Tensor]]:
"""
Returns the callback function to load images for SSD.
@ -301,43 +296,30 @@ def _load_images_ssd_callback(resized_shape: Sequence[int]) \
callback function
"""
def _load_images_ssd(paths: Sequence[str],
labels: Sequence[Sequence[Tuple[int, int, int, int, int]]]) \
-> Tuple[tf.Tensor, Sequence[Sequence[Tuple[int, int, int, int, int]]]]:
def _load_images_ssd(paths: Sequence[str]) -> Tuple[tf.Tensor]:
"""
Callback function to load images for SSD.
Args:
paths: paths to the images
labels: ground truth data for the images
Returns:
loaded images
"""
_images = tf.map_fn(lambda path: tf.read_file(path), paths)
def _get_images(image_data: Sequence[tf.Tensor]) -> List[tf.Tensor]:
image = tf.image.decode_image(image_data[0], channels=3, dtype=tf.float32)
def _get_images(image_data: tf.Tensor) -> List[tf.Tensor]:
image = tf.image.decode_image(image_data, channels=3, dtype=tf.float32)
image_shape = tf.shape(image)
image = tf.reshape(image, [image_shape[0], image_shape[1], 3])
label = image_data[1]
xmin = label[1]
ymin = label[2]
xmax = label[3]
ymax = label[4]
image_resized = tf.image.resize(image, resized_shape[0], resized_shape[1])
# also resize labels
processed_label = label[:]
processed_label[:, [xmin, xmax]] = tf.round(label[:, [xmin, xmax]] * (resized_shape[0] / image_shape[0]))
processed_label[:, [ymin, ymax]] = tf.round(label[:, [ymin, ymax]] * (resized_shape[1] / image_shape[1]))
return [image_resized, processed_label]
return image_resized
processed = tf.map_fn(_get_images, [_images, labels], dtype=[tf.float32, tf.float32])
processed_images = processed[0]
processed_images = tf.reshape(processed_images, [-1, resized_shape[0], resized_shape[1], 3])
processed = tf.map_fn(_get_images, _images, dtype=[tf.float32, tf.float32])
processed_images = tf.reshape(processed, [-1, resized_shape[0], resized_shape[1], 3])
return processed_images, processed[1]
return processed_images
return _load_images_ssd

2
src/twomartens/masterthesis/ssd.py
View File

@ -173,7 +173,7 @@ def _predict_one_epoch(dataset: tf.data.Dataset,
# go through the data set
counter = 0
for inputs, _ in dataset:
for inputs in dataset:
decoded_predictions_batch = []
if use_dropout:
for _ in range(forward_passes_per_image):