def train():
"""The main function that runs training"""
## data
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = \
datasets.get_dataset(FLAGS.dataset_name,
FLAGS.dataset_split_name,
FLAGS.dataset_dir,
FLAGS.im_batch,
is_training=True)
data_queue = tf.RandomShuffleQueue(
capacity=32,
min_after_dequeue=16,
dtypes=(image.dtype, ih.dtype, iw.dtype, gt_boxes.dtype,
gt_masks.dtype, num_instances.dtype, img_id.dtype))
enqueue_op = data_queue.enqueue(
(image, ih, iw, gt_boxes, gt_masks, num_instances, img_id))
data_queue_runner = tf.train.QueueRunner(data_queue, [enqueue_op] * 4)
tf.add_to_collection(tf.GraphKeys.QUEUE_RUNNERS, data_queue_runner)
(image, ih, iw, gt_boxes, gt_masks, num_instances,
img_id) = data_queue.dequeue()
im_shape = tf.shape(image)
image = tf.reshape(image, (im_shape[0], im_shape[1], im_shape[2], 3))
## network
logits, end_points, pyramid_map = network.get_network(
FLAGS.network,
image,
weight_decay=FLAGS.weight_decay,
is_training=True)
outputs = pyramid_network.build(end_points,
im_shape[1],
im_shape[2],
pyramid_map,
num_classes=81,
base_anchors=9,
is_training=True,
gt_boxes=gt_boxes,
gt_masks=gt_masks,
loss_weights=[0.2, 0.2, 1.0, 0.2, 1.0])
total_loss = outputs['total_loss']
losses = outputs['losses']
batch_info = outputs['batch_info']
regular_loss = tf.add_n(
tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
input_image = end_points['input']
final_box = outputs['final_boxes']['box']
final_cls = outputs['final_boxes']['cls']
final_prob = outputs['final_boxes']['prob']
final_gt_cls = outputs['final_boxes']['gt_cls']
# Get the computation result from output
final_mask = outputs['mask']['mask']
print('final_mask', final_mask.shape)
gt = outputs['gt']
# replace the draw_bbox
# draw_mask(step,
# np.uint8((np.array(input_imagenp[0])/2.0+0.5)*255.0),
# name='est',
# bbox=final_boxnp,
# mask=final_masknp,
# label=final_clsnp,
# prob=final_probnp,
# gt_label=np.argmax(np.asarray(final_gt_clsnp),axis=1),
# )
# draw_mask(step,
# np.uint8((np.array(input_imagenp[0])/2.0+0.5)*255.0),
# name='gt',
# bbox=gtnp[:,0:4],
# mask=gt_masksnp,
# label=np.asarray(gtnp[:,4], dtype=np.uint8),
# )
#############################
tmp_0 = outputs['losses']
tmp_1 = outputs['losses']
tmp_2 = outputs['losses']
tmp_3 = outputs['losses']
tmp_4 = outputs['losses']
# tmp_0 = outputs['tmp_0']
# tmp_1 = outputs['tmp_1']
# tmp_2 = outputs['tmp_2']
tmp_3 = outputs['tmp_3']
tmp_4 = outputs['tmp_4']
############################
## solvers
global_step = slim.create_global_step()
update_op = solve(global_step)
cropped_rois = tf.get_collection('__CROPPED__')[0]
transposed = tf.get_collection('__TRANSPOSED__')[0]
#os.environ["CUDA_VISIBLE_DEVICES"]="1"
config = tf.ConfigProto(allow_soft_placement=True)
#config.gpu_options.allow_growth = True
config.gpu_options.per_process_gpu_memory_fraction = 0.95
##gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.45)
##sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
sess = tf.Session(config=config)
init_op = tf.group(tf.global_variables_initializer(),
tf.local_variables_initializer())
sess.run(init_op)
summary_op = tf.summary.merge_all()
logdir = os.path.join(FLAGS.train_dir, strftime('%Y%m%d%H%M%S', gmtime()))
if not os.path.exists(logdir):
os.makedirs(logdir)
summary_writer = tf.summary.FileWriter(logdir, graph=sess.graph)
## restore
restore(sess)
## main loop
coord = tf.train.Coordinator()
threads = []
# print (tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS))
for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS):
threads.extend(
qr.create_threads(sess, coord=coord, daemon=True, start=True))
tf.train.start_queue_runners(sess=sess, coord=coord)
saver = tf.train.Saver(max_to_keep=20)
for step in range(FLAGS.max_iters):
start_time = time.time()
#rGet the mask results
s_, tot_loss, reg_lossnp, img_id_str, \
rpn_box_loss, rpn_cls_loss, refined_box_loss, refined_cls_loss, mask_loss, \
gt_boxesnp, \
rpn_batch_pos, rpn_batch, refine_batch_pos, refine_batch, mask_batch_pos, mask_batch, \
input_imagenp, final_masknp, final_boxnp, final_clsnp, final_probnp, final_gt_clsnp, gtnp,gt_masksnp, tmp_0np, tmp_1np, tmp_2np, tmp_3np, tmp_4np= \
sess.run([update_op, total_loss, regular_loss, img_id] +
losses +
[gt_boxes] +
batch_info +
[input_image] + [final_mask] + [final_box] + [final_cls] + [final_prob] + [final_gt_cls] + [gt] + [gt_masks] + [tmp_0] + [tmp_1] + [tmp_2] + [tmp_3] + [tmp_4])
duration_time = time.time() - start_time
if step % 1 == 0:
print(
"""iter %d: image-id:%07d, time:%.3f(sec), regular_loss: %.6f, """
"""total-loss %.4f(%.4f, %.4f, %.6f, %.4f, %.4f), """
"""instances: %d, """
"""batch:(%d|%d, %d|%d, %d|%d)""" %
(step, img_id_str, duration_time, reg_lossnp, tot_loss,
rpn_box_loss, rpn_cls_loss, refined_box_loss,
refined_cls_loss, mask_loss, gt_boxesnp.shape[0],
rpn_batch_pos, rpn_batch, refine_batch_pos, refine_batch,
mask_batch_pos, mask_batch))
draw_bbox(
step,
np.uint8((np.array(input_imagenp[0]) / 2.0 + 0.5) * 255.0),
name='est',
bbox=final_boxnp,
label=final_clsnp,
prob=final_probnp,
gt_label=np.argmax(np.asarray(final_gt_clsnp), axis=1),
)
draw_bbox(
step,
np.uint8((np.array(input_imagenp[0]) / 2.0 + 0.5) * 255.0),
name='gt',
bbox=gtnp[:, 0:4],
label=np.asarray(gtnp[:, 4], dtype=np.uint8),
)
print("labels")
# print (cat_id_to_cls_name(np.unique(np.argmax(np.asarray(final_gt_clsnp),axis=1)))[1:])
# print (cat_id_to_cls_name(np.unique(np.asarray(gt_boxesnp, dtype=np.uint8)[:,4])))
print(
cat_id_to_cls_name(
np.unique(np.argmax(np.asarray(tmp_3np), axis=1)))[1:])
#print (cat_id_to_cls_name(np.unique(np.argmax(np.asarray(gt_boxesnp)[:,4],axis=1))))
print("classes")
print(
cat_id_to_cls_name(
np.unique(np.argmax(np.array(tmp_4np), axis=1))))
# print (np.asanyarray(tmp_3np))
#print ("ordered rois")
#print (np.asarray(tmp_0np)[0])
#print ("pyramid_feature")
#print ()
#print(np.unique(np.argmax(np.array(final_probnp),axis=1)))
#for var, val in zip(tmp_2, tmp_2np):
# print(var.name)
#print(np.argmax(np.array(tmp_0np),axis=1))
if np.isnan(tot_loss) or np.isinf(tot_loss):
print(gt_boxesnp)
raise
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
summary_writer.flush()
if (step % 10000 == 0 or step + 1 == FLAGS.max_iters) and step != 0:
checkpoint_path = os.path.join(
FLAGS.train_dir,
FLAGS.dataset_name + '_' + FLAGS.network + '_model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
if coord.should_stop():
coord.request_stop()
coord.join(threads)
def test():
"""The main function that runs training"""
## data
image, original_image_height, original_image_width, image_height, image_width, gt_boxes, gt_masks, num_instances, image_id = \
datasets.get_dataset(FLAGS.dataset_name,
FLAGS.dataset_split_name_test,
FLAGS.dataset_dir,
FLAGS.im_batch,
is_training=False)
im_shape = tf.shape(image)
image = tf.reshape(image, (im_shape[0], im_shape[1], im_shape[2], 3))
## network
logits, end_points, pyramid_map = network.get_network(FLAGS.network, image,
weight_decay=0.0, batch_norm_decay=0.0, is_training=True)
outputs = pyramid_network.build(end_points, im_shape[1], im_shape[2], pyramid_map,
num_classes=81,
base_anchors=3,
is_training=False,
gt_boxes=None, gt_masks=None, loss_weights=[0.0, 0.0, 0.0, 0.0, 0.0])
input_image = end_points['input']
testing_mask_rois = outputs['mask_ordered_rois']
testing_mask_final_mask = outputs['mask_final_mask']
testing_mask_final_clses = outputs['mask_final_clses']
testing_mask_final_scores = outputs['mask_final_scores']
## solvers
global_step = slim.create_global_step()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.8)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
# init_op = tf.group(
# tf.global_variables_initializer(),
# tf.local_variables_initializer()
# )
# sess.run(init_op)
# summary_op = tf.summary.merge_all()
logdir = os.path.join(FLAGS.train_dir, strftime('%Y%m%d%H%M%S', gmtime()))
if not os.path.exists(logdir):
os.makedirs(logdir)
summary_writer = tf.summary.FileWriter(logdir, graph=sess.graph)
## restore
restore(sess)
tf.train.start_queue_runners(sess=sess)
## main loop
# for step in range(FLAGS.max_iters):
for step in range(82783):#range(40503):
start_time = time.time()
image_id_str, original_image_heightnp, original_image_widthnp, image_heightnp, image_widthnp, \
gt_boxesnp, gt_masksnp,\
input_imagenp,\
testing_mask_roisnp, testing_mask_final_masknp, testing_mask_final_clsesnp, testing_mask_final_scoresnp = \
sess.run([image_id] + [original_image_height] + [original_image_width] + [image_height] + [image_width] +\
[gt_boxes] + [gt_masks] +\
[input_image] + \
[testing_mask_rois] + [testing_mask_final_mask] + [testing_mask_final_clses] + [testing_mask_final_scores])
duration_time = time.time() - start_time
if step % 1 == 0:
print ( """iter %d: image-id:%07d, time:%.3f(sec), """
"""instances: %d, """
% (step, image_id_str, duration_time,
gt_boxesnp.shape[0]))
if step % 1 == 0:
draw_bbox(step,
np.uint8((np.array(input_imagenp[0])/2.0+0.5)*255.0),
name='test_est',
bbox=testing_mask_roisnp,
label=testing_mask_final_clsesnp,
prob=testing_mask_final_scoresnp,
mask=testing_mask_final_masknp,
vis_th=0.5)
draw_bbox(step,
np.uint8((np.array(input_imagenp[0])/2.0+0.5)*255.0),
name='test_gt',
bbox=gt_boxesnp[:,0:4],
label=gt_boxesnp[:,4].astype(np.int32),
prob=np.ones((gt_boxesnp.shape[0],81), dtype=np.float32),)
print ("predict")
# LOG (cat_id_to_cls_name(np.unique(np.argmax(np.array(training_rcnn_clsesnp),axis=1))))
print (cat_id_to_cls_name(testing_mask_final_clsesnp))
print (np.max(np.array(testing_mask_final_scoresnp),axis=1))
_collectData(image_id_str, testing_mask_final_clsesnp, testing_mask_roisnp, testing_mask_final_scoresnp, original_image_heightnp, original_image_widthnp, image_heightnp, image_widthnp, testing_mask_final_masknp)
def train():
"""The main function that runs training"""
## data
image, original_image_height, original_image_width, image_height, image_width, gt_boxes, gt_masks, num_instances, image_id = \
datasets.get_dataset(FLAGS.dataset_name,
FLAGS.dataset_split_name,
FLAGS.dataset_dir,
FLAGS.im_batch,
is_training=True)
## queuing data
data_queue = tf.RandomShuffleQueue(capacity=32, min_after_dequeue=16,
dtypes=(
image.dtype, original_image_height.dtype, original_image_width.dtype, image_height.dtype, image_width.dtype,
gt_boxes.dtype, gt_masks.dtype,
num_instances.dtype, image_id.dtype))
enqueue_op = data_queue.enqueue((image, original_image_height, original_image_width, image_height, image_width, gt_boxes, gt_masks, num_instances, image_id))
data_queue_runner = tf.train.QueueRunner(data_queue, [enqueue_op] * 4)
tf.add_to_collection(tf.GraphKeys.QUEUE_RUNNERS, data_queue_runner)
(image, original_image_height, original_image_width, image_height, image_width, gt_boxes, gt_masks, num_instances, image_id) = data_queue.dequeue()
im_shape = tf.shape(image)
image = tf.reshape(image, (im_shape[0], im_shape[1], im_shape[2], 3))
## network
logits, end_points, pyramid_map = network.get_network(FLAGS.network, image,
weight_decay=FLAGS.weight_decay, batch_norm_decay=FLAGS.batch_norm_decay, is_training=True)
outputs = pyramid_network.build(end_points, image_height, image_width, pyramid_map,
num_classes=81,
base_anchors=3,#9#15
is_training=True,
gt_boxes=gt_boxes, gt_masks=gt_masks,
loss_weights=[1.0, 1.0, 10.0, 1.0, 10.0])
# loss_weights=[10.0, 1.0, 0.0, 0.0, 0.0])
# loss_weights=[100.0, 100.0, 1000.0, 10.0, 100.0])
# loss_weights=[0.2, 0.2, 1.0, 0.2, 1.0])
# loss_weights=[0.1, 0.01, 10.0, 0.1, 1.0])
total_loss = outputs['total_loss']
losses = outputs['losses']
batch_info = outputs['batch_info']
regular_loss = tf.add_n(tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
input_image = end_points['input']
training_rcnn_rois = outputs['training_rcnn_rois']
training_rcnn_clses = outputs['training_rcnn_clses']
training_rcnn_clses_target = outputs['training_rcnn_clses_target']
training_rcnn_scores = outputs['training_rcnn_scores']
training_mask_rois = outputs['training_mask_rois']
training_mask_clses_target = outputs['training_mask_clses_target']
training_mask_final_mask = outputs['training_mask_final_mask']
training_mask_final_mask_target = outputs['training_mask_final_mask_target']
tmp_0 = outputs['rpn']['P2']['shape']
tmp_1 = outputs['rpn']['P3']['shape']
tmp_2 = outputs['rpn']['P4']['shape']
tmp_3 = outputs['rpn']['P5']['shape']
## solvers
global_step = slim.create_global_step()
update_op = solve(global_step)
cropped_rois = tf.get_collection('__CROPPED__')[0]
transposed = tf.get_collection('__TRANSPOSED__')[0]
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.95)
#gpu_options = tf.GPUOptions(allow_growth=True)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
#sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))
init_op = tf.group(
tf.global_variables_initializer(),
tf.local_variables_initializer()
)
sess.run(init_op)
summary_op = tf.summary.merge_all()
logdir = os.path.join(FLAGS.train_dir, strftime('%Y%m%d%H%M%S', gmtime()))
if not os.path.exists(logdir):
os.makedirs(logdir)
summary_writer = tf.summary.FileWriter(logdir, graph=sess.graph)
## restore
restore(sess)
## coord settings
coord = tf.train.Coordinator()
threads = []
for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS):
threads.extend(qr.create_threads(sess, coord=coord, daemon=True,
start=True))
tf.train.start_queue_runners(sess=sess, coord=coord)
## saver init
saver = tf.train.Saver(max_to_keep=20)
## finalize the graph for checking memory leak
sess.graph.finalize()
## main loop
for step in range(FLAGS.max_iters):
start_time = time.time()
s_, tot_loss, reg_lossnp, image_id_str, \
rpn_box_loss, rpn_cls_loss, rcnn_box_loss, rcnn_cls_loss, mask_loss, \
gt_boxesnp, tmp_0np, tmp_1np, tmp_2np, tmp_3np, \
rpn_batch_pos, rpn_batch, rcnn_batch_pos, rcnn_batch, mask_batch_pos, mask_batch, \
input_imagenp, \
training_rcnn_roisnp, training_rcnn_clsesnp, training_rcnn_clses_targetnp, training_rcnn_scoresnp, training_mask_roisnp, training_mask_clses_targetnp, training_mask_final_masknp, training_mask_final_mask_targetnp = \
sess.run([update_op, total_loss, regular_loss, image_id] +
losses +
[gt_boxes] + [tmp_0] + [tmp_1] + [tmp_2] +[tmp_3] +
batch_info +
[input_image] +
[training_rcnn_rois] + [training_rcnn_clses] + [training_rcnn_clses_target] + [training_rcnn_scores] + [training_mask_rois] + [training_mask_clses_target] + [training_mask_final_mask] + [training_mask_final_mask_target])
duration_time = time.time() - start_time
if step % 1 == 0:
LOG ( """iter %d: image-id:%07d, time:%.3f(sec), regular_loss: %.6f, """
"""total-loss %.4f(%.4f, %.4f, %.6f, %.4f, %.4f), """
"""instances: %d, """
"""batch:(%d|%d, %d|%d, %d|%d)"""
% (step, image_id_str, duration_time, reg_lossnp,
tot_loss, rpn_box_loss, rpn_cls_loss, rcnn_box_loss, rcnn_cls_loss, mask_loss,
gt_boxesnp.shape[0],
rpn_batch_pos, rpn_batch, rcnn_batch_pos, rcnn_batch, mask_batch_pos, mask_batch))
LOG ("target")
LOG (cat_id_to_cls_name(np.unique(np.argmax(np.asarray(training_rcnn_clses_targetnp),axis=1))))
LOG ("predict")
LOG (cat_id_to_cls_name(np.unique(np.argmax(np.array(training_rcnn_clsesnp),axis=1))))
LOG (tmp_0np)
LOG (tmp_1np)
LOG (tmp_2np)
LOG (tmp_3np)
if step % 50 == 0:
draw_bbox(step,
np.uint8((np.array(input_imagenp[0])/2.0+0.5)*255.0),
name='train_est',
bbox=training_rcnn_roisnp,
label=np.argmax(np.array(training_rcnn_scoresnp),axis=1),
prob=training_rcnn_scoresnp,
# bbox=training_mask_roisnp,
# label=training_mask_clses_targetnp,
# prob=np.zeros((training_mask_final_masknp.shape[0],81), dtype=np.float32)+1.0,
# mask=training_mask_final_masknp,
vis_all=True)
draw_bbox(step,
np.uint8((np.array(input_imagenp[0])/2.0+0.5)*255.0),
name='train_gt',
bbox=training_rcnn_roisnp,
label=np.argmax(np.array(training_rcnn_clses_targetnp),axis=1),
prob=np.zeros((training_rcnn_clsesnp.shape[0],81), dtype=np.float32)+1.0,
# bbox=training_mask_roisnp,
# label=training_mask_clses_targetnp,
# prob=np.zeros((training_mask_final_masknp.shape[0],81), dtype=np.float32)+1.0,
# mask=training_mask_final_mask_targetnp,
vis_all=True)
if np.isnan(tot_loss) or np.isinf(tot_loss):
LOG (gt_boxesnp)
raise
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
summary_writer.flush()
if (step % 500 == 0 or step + 1 == FLAGS.max_iters) and step != 0:
checkpoint_path = os.path.join(FLAGS.train_dir,
FLAGS.dataset_name + '_' + FLAGS.network + '_model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
if coord.should_stop():
coord.request_stop()
coord.join(threads)
gc.collect()
def test():
"""The main function that runs training"""
## data
image, ih, iw, gt_boxes, gt_masks, num_instances, img_id = \
datasets.get_dataset(FLAGS.dataset_name,
FLAGS.dataset_split_name,
FLAGS.dataset_dir,
FLAGS.im_batch,
is_training=False)
im_shape = tf.shape(image)
image = tf.reshape(image, (im_shape[0], im_shape[1], im_shape[2], 3))
## network
logits, end_points, pyramid_map = network.get_network(
FLAGS.network,
image,
weight_decay=FLAGS.weight_decay,
is_training=False)
outputs = pyramid_network.build(end_points,
im_shape[1],
im_shape[2],
pyramid_map,
num_classes=81,
base_anchors=15,
is_training=False,
gt_boxes=None,
gt_masks=None,
loss_weights=[0.0, 0.0, 0.0, 0.0, 0.0])
input_image = end_points['input']
testing_mask_rois = outputs['mask_ordered_rois']
testing_mask_final_mask = outputs['mask_final_mask']
testing_mask_final_clses = outputs['mask_final_clses']
testing_mask_final_scores = outputs['mask_final_scores']
#############################
tmp_0 = outputs['tmp_0']
tmp_1 = outputs['tmp_1']
tmp_2 = outputs['tmp_2']
tmp_3 = outputs['tmp_3']
tmp_4 = outputs['tmp_4']
tmp_5 = outputs['tmp_5']
############################
## solvers
global_step = slim.create_global_step()
#update_op = solve(global_step)
cropped_rois = tf.get_collection('__CROPPED__')[0]
transposed = tf.get_collection('__TRANSPOSED__')[0]
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.8)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
init_op = tf.group(tf.global_variables_initializer(),
tf.local_variables_initializer())
sess.run(init_op)
summary_op = tf.summary.merge_all()
logdir = os.path.join(FLAGS.train_dir, strftime('%Y%m%d%H%M%S', gmtime()))
if not os.path.exists(logdir):
os.makedirs(logdir)
summary_writer = tf.summary.FileWriter(logdir, graph=sess.graph)
## restore
restore(sess)
## main loop
coord = tf.train.Coordinator()
threads = []
# print (tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS))
for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS):
threads.extend(
qr.create_threads(sess, coord=coord, daemon=True, start=True))
tf.train.start_queue_runners(sess=sess, coord=coord)
saver = tf.train.Saver(max_to_keep=20)
for step in range(FLAGS.max_iters):
start_time = time.time()
img_id_str, \
gt_boxesnp, \
input_imagenp, tmp_0np, tmp_1np, tmp_2np, tmp_3np, tmp_4np, tmp_5np, \
testing_mask_roisnp, testing_mask_final_masknp, testing_mask_final_clsesnp, testing_mask_final_scoresnp = \
sess.run([img_id] + \
[gt_boxes] + \
[input_image] + [tmp_0] + [tmp_1] + [tmp_2] + [tmp_3] + [tmp_4] + [tmp_5] + \
[testing_mask_rois] + [testing_mask_final_mask] + [testing_mask_final_clses] + [testing_mask_final_scores])
duration_time = time.time() - start_time
if step % 1 == 0:
print("""iter %d: image-id:%07d, time:%.3f(sec), """
"""instances: %d, """ %
(step, img_id_str, duration_time, gt_boxesnp.shape[0]))
if step % 1 == 0:
draw_bbox(
step,
np.uint8((np.array(input_imagenp[0]) / 2.0 + 0.5) * 255.0),
name='test_est',
bbox=testing_mask_roisnp,
label=testing_mask_final_clsesnp,
prob=testing_mask_final_scoresnp,
mask=testing_mask_final_masknp,
)
def forward_test_single_image():
if not os.path.exists(save_dir_bbox):
os.makedirs(save_dir_bbox)
if not os.path.exists(save_dir_mask):
os.makedirs(save_dir_mask)
file_pathname = testdata_base_dir + '*.' + file_pattern
image_paths = glob.glob(file_pathname) #with .jpg/.png
image_names = glob.glob(file_pathname) #no .jpg/.png
for i in range(len(image_paths)):
image_names[i] = image_paths[i][len(testdata_base_dir):-4]
print(image_paths)
print(image_names)
TEST_image = tf.placeholder(tf.float32, shape=[1, None, None, 3])
im_shape = tf.shape(TEST_image)
## network
logits, end_points, pyramid_map = network.get_network(
FLAGS.network,
TEST_image,
weight_decay=FLAGS.weight_decay,
is_training=True)
outputs = pyramid_network.build(end_points,
im_shape[1],
im_shape[2],
pyramid_map,
num_classes=81,
base_anchors=9,
is_training=True,
gt_boxes=None,
gt_masks=None,
loss_weights=[0.2, 0.2, 1.0, 0.2, 1.0])
input_image = end_points['input']
print("input_image.shape", input_image.shape)
final_box = outputs['final_boxes']['box']
final_cls = outputs['final_boxes']['cls']
final_prob = outputs['final_boxes']['prob']
final_mask = outputs['mask']['mask']
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
# config.gpu_options.per_process_gpu_memory_fraction = 0.5
sess = tf.Session(config=config)
init_op = tf.group(tf.global_variables_initializer(),
tf.local_variables_initializer())
sess.run(init_op)
## restore trained model
restore(sess)
## main loop
coord = tf.train.Coordinator()
threads = []
# print (tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS))
for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS):
threads.extend(
qr.create_threads(sess, coord=coord, daemon=True, start=True))
tf.train.start_queue_runners(sess=sess, coord=coord)
for i in range(len(image_paths)):
## read test image
test_img = Image.open(image_paths[i])
test_img = test_img.convert("RGB")
test_img_i = np.array(test_img, dtype=np.uint8)
test_img = np.array(test_img, dtype=np.float32)
test_img = test_img[np.newaxis, ...]
print("test_img.shape", test_img.shape)
print("test_img_i.shape", test_img_i.shape)
# start_time = time.time()
input_imagenp, final_boxnp, final_clsnp, final_probnp, \
final_masknp= \
sess.run([input_image] + [final_box] + [final_cls] + [final_prob] +
[final_mask], feed_dict={TEST_image:test_img})
# duration_time = time.time() - start_time
draw_bbox(
test_img_i,
type='est',
bbox=final_boxnp,
label=final_clsnp,
prob=final_probnp,
gt_label=None,
save_dir=save_dir_bbox,
save_name=image_names[i],
)
print("final_masknp.shape\n", final_masknp.shape)
draw_mask(
test_img_i,
type='est',
bbox=final_boxnp,
mask=final_masknp,
label=final_clsnp,
prob=final_probnp,
gt_label=None,
save_dir=save_dir_mask,
save_name=image_names[i],
)
if coord.should_stop():
coord.request_stop()
coord.join(threads)