def pack_model(inputs, train = True, back_network = 'resnet50',
#num_classes=81,
num_classes=91,
base_anchors=9,
weight_decay=0.00005,
**kwargs
):
# Reshape the input image, batch size 1 supported
image = inputs['images']
ih = inputs['height']
iw = inputs['width']
im_shape = tf.shape(image)
#image = tf.Print(image, [im_shape], message = 'shape', summarize = 4)
image = tf.reshape(image, (im_shape[0], im_shape[1], im_shape[2], 3))
image = tf.cast(image, tf.float32)
image_height = ih
image_width = iw
num_instances = inputs['num_objects']
gt_boxes = inputs['bboxes']
#gt_boxes = tf.reshape(gt_boxes, [num_instances, 4])
#labels = inputs['labels']
#labels = tf.reshape(labels, [num_instances, 1])
#gt_boxes = tf.concat([gt_boxes, tf.cast(labels, tf.float64)], 1)
#gt_boxes = tf.Print(gt_boxes, [tf.shape(gt_boxes)], message = 'Box shape', summarize = 4)
# Build the basic network
logits, end_points, pyramid_map = network.get_network(back_network, image,
weight_decay=weight_decay)
# Build the pyramid
pyramid = pyramid_network.build_pyramid(pyramid_map, end_points)
# Build the heads
outputs = \
pyramid_network.build_heads(pyramid, image_height, image_width, num_classes, base_anchors,
is_training=train, gt_boxes=gt_boxes)
return {'outputs': outputs, 'pyramid': pyramid}, {'network': back_network}
## 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:
summaries.add(tf.summary.histogram('pyramid/hist/' + p,
pyramid[p]))
summaries.add(
tf.summary.scalar('pyramid/means/' + p,
tf.reduce_mean(tf.abs(pyramid[p]))))
outputs = pyramid_network.build_heads(pyramid,
ih,
iw,
num_classes=81,