def get_dataset(dataset_name,
split_name,
dataset_dir,
im_batch=1,
is_training=False,
file_pattern=None,
reader=None):
""""""
if file_pattern is None:
file_pattern = dataset_name + '_' + split_name + '*.tfrecord'
tfrecords = glob.glob(dataset_dir + '/records/' + file_pattern)
assert len(
tfrecords
) > 0, "haven't found any tfrecord(did you run train.py from code root?). we were looking at %s." % dataset_dir + '/records/' + file_pattern
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = coco.read(
tfrecords, is_training=is_training)
image, new_ih, new_iw, gt_boxes, gt_masks = coco_preprocess.preprocess_image(
image, gt_boxes, gt_masks, is_training)
#visualize_input(gt_boxes, image, tf.expand_dims(gt_masks, axis=3))
return image, ih, iw, new_ih, new_iw, gt_boxes, gt_masks, num_instances, img_id
def get_dataset(dataset_name, split_name, dataset_dir,
im_batch=1, is_training=False, file_pattern=None, reader=None):
""""""
if file_pattern is None:
file_pattern = dataset_name + '_' + split_name + '*.tfrecord'
tfrecords = glob.glob(dataset_dir + '/records/' + file_pattern)
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = coco.read(tfrecords)
image, gt_boxes, gt_masks = coco_preprocess.preprocess_image(image, gt_boxes, gt_masks, is_training)
return image, ih, iw, gt_boxes, gt_masks, num_instances, img_id
def prep_data(self, data):
for i in range(len(data)):
inputs = data[i]
image = inputs['images']
image = tf.decode_raw(image, tf.uint8)
ih = inputs['height']
iw = inputs['width']
ih = tf.cast(ih, tf.int32)
iw = tf.cast(iw, tf.int32)
inputs['height'] = ih
inputs['width'] = iw
imsize = tf.size(image)
#image = tf.Print(image, [imsize, ih, iw], message = 'Imsize')
image = tf.cond(tf.equal(imsize, ih * iw), \
lambda: tf.image.grayscale_to_rgb(tf.reshape(image, (ih, iw, 1))), \
lambda: tf.reshape(image, (ih, iw, 3)))
image_height = ih
image_width = iw
num_instances = inputs['num_objects']
num_instances = tf.cast(num_instances, tf.int32)
inputs['num_objects'] = num_instances
gt_boxes = tf.decode_raw(inputs['bboxes'], tf.float64)
gt_boxes = tf.reshape(gt_boxes, [num_instances, 4])
labels = tf.decode_raw(inputs['labels'], tf.int32)
labels = tf.reshape(labels, [num_instances, 1])
inputs['labels'] = labels
gt_boxes = tf.concat([gt_boxes, tf.cast(labels, tf.float64)], 1)
gt_boxes = tf.cast(gt_boxes, tf.float32)
gt_masks = tf.decode_raw(inputs['segmentation_masks'], tf.uint8)
gt_masks = tf.cast(gt_masks, tf.int32)
gt_masks = tf.reshape(gt_masks, [num_instances, ih, iw])
#gt_masks = tf.Print(gt_masks, [tf.shape(gt_masks)], message = 'Mask shape before', summarize = 4)
image, gt_boxes, gt_masks = coco_preprocess.preprocess_image(image, gt_boxes, gt_masks, True)
#image = tf.Print(image, [tf.shape(image)], message = 'Imsize', summarize = 4)
#gt_masks = tf.Print(gt_masks, [tf.shape(gt_masks)], message = 'Mask shape', summarize = 4)
inputs['segmentation_masks'] = gt_masks
inputs['bboxes'] = gt_boxes
inputs['images'] = image
return data
def get_dataset(dataset_name,
split_name,
dataset_dir,
im_batch=1,
is_training=False,
file_pattern=None,
reader=None):
""""""
if file_pattern is None:
file_pattern = dataset_name + '_' + split_name + '*.tfrecord'
pattern = '../' + dataset_dir + 'records/' + file_pattern
tfrecords = glob.glob(pattern)
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = coco.read(
tfrecords)
image, gt_boxes, gt_masks = coco_preprocess.preprocess_image(
image, gt_boxes, gt_masks, is_training)
#visualize_input(gt_boxes, image, tf.expand_dims(gt_masks, axis=3))
return image, ih, iw, gt_boxes, gt_masks, num_instances, img_id
DEBUG = False
with tf.Graph().as_default():
gpu_options = tf.GPUOptions(
per_process_gpu_memory_fraction=0.8,
allow_growth=True,
)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options,
allow_soft_placement=True)) as sess:
global_step = slim.create_global_step()
## data
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = \
coco.read('./data/coco/records/coco_train2014_00000-of-00040.tfrecord')
with tf.control_dependencies([image, gt_boxes, gt_masks]):
image, gt_boxes, gt_masks = coco_preprocess.preprocess_image(
image, gt_boxes, gt_masks, is_training=True)
## network
with slim.arg_scope(resnet_v1.resnet_arg_scope(weight_decay=0.0001)):
logits, end_points = resnet50(image, 1000, is_training=False)
end_points['inputs'] = image
for x in sorted(end_points.keys()):
print(x, end_points[x].name, end_points[x].shape)
pyramid = pyramid_network.build_pyramid('resnet50', end_points)
# for p in pyramid:
# print (p, pyramid[p])
summaries = set(tf.get_collection(tf.GraphKeys.SUMMARIES))
for p in pyramid:
import libs.nets.resnet_v1 as resnet_v1
import libs.datasets.dataset_factory as dataset_factory
import libs.datasets.coco as coco
import libs.preprocessings.coco_v1 as preprocess_coco
from libs.layers import ROIAlign
resnet50 = resnet_v1.resnet_v1_50
FLAGS = tf.app.flags.FLAGS
with tf.Graph().as_default():
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = \
coco.read('./data/coco/records/coco_trainval2014_00000-of-00048.tfrecord')
image, gt_boxes, gt_masks = \
preprocess_coco.preprocess_image(image, gt_boxes, gt_masks)
sess = tf.Session()
init_op = tf.group(tf.global_variables_initializer(),
tf.local_variables_initializer())
# init_op = tf.initialize_all_variables()
boxes = [[100, 100, 200, 200],
[50, 50, 100, 100],
[100, 100, 750, 750],
[50, 50, 60, 60]]
# boxes = np.zeros((0, 4))
boxes = tf.constant(boxes, tf.float32)
feat = ROIAlign(image, boxes, False, 16, 7, 7)
return tfrecords
with tf.Graph().as_default():
records = get_records(FLAGS.dataset_name,
FLAGS.dataset_split_name,
FLAGS.dataset_dir,
FLAGS.im_batch,
is_training=False)
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = \
coco.read(records)
image, gt_boxes, gt_masks = \
preprocess_coco.preprocess_image(image, gt_boxes, gt_masks)
# using queue to input
queue = tf.RandomShuffleQueue(capacity=12,
min_after_dequeue=6,
dtypes=(image.dtype, ih.dtype, iw.dtype,
gt_boxes.dtype, gt_masks.dtype,
num_instances.dtype, img_id.dtype))
enqueue_op = queue.enqueue(
(image, ih, iw, gt_boxes, gt_masks, num_instances, img_id))
(image_, ih_, iw_, gt_boxes_, gt_masks_, num_instances_,
img_id_) = queue.dequeue()
qr = tf.train.QueueRunner(queue, [enqueue_op] * 4)
sess = tf.Session()
init_op = tf.group(tf.global_variables_initializer(),
cls = np.random.randint(1, 6, (N, 1)).astype(np.float32)
gt_boxes = np.hstack((xy, xy + wh, cls)).astype(np.float32)
gt_boxes_np = gt_boxes
image_np = image
gt_masks_np = gt_masks
for i in range(N):
box = gt_boxes[i, 0:4]
gt_masks[i, int(box[1]):int(box[3]),
int(box[0]):int(box[2])] = 1
image = tf.constant(image)
gt_boxes = tf.constant(gt_boxes)
gt_masks = tf.constant(gt_masks)
image, gt_boxes, gt_masks = \
coco_preprocess.preprocess_image(image, gt_boxes, gt_masks, is_training=True)
with tf.Session() as sess:
# print(image.eval())
image_tf, gt_boxes_tf, gt_masks_tf = \
sess.run([image, gt_boxes, gt_masks])
print ('#######################')
print ('DATA PREPROCESSING TEST')
print ('#######################')
print ('gt_boxes shape:', gt_boxes_tf.shape)
print('mask shape:', gt_masks_tf.shape)
print(gt_boxes_tf)
for i in range(N):
box = np.round(gt_boxes_tf[i, 0:4])
box = box.astype(np.int32)
m = gt_masks_tf[i, box[1]:box[3], box[0]:box[2]]