def save_image_with_bbox(image, labels_, scores_, bboxes_):
if not hasattr(save_image_with_bbox, "counter"):
save_image_with_bbox.counter = 0 # it doesn't exist yet, so initialize it
save_image_with_bbox.counter += 1
img_to_draw = np.copy(image).astype(np.uint8)
img_to_draw = draw_toolbox.bboxes_draw_on_img(img_to_draw, labels_, scores_, bboxes_, thickness=2)
imsave(os.path.join('./debug/{}.jpg').format(save_image_with_bbox.counter), img_to_draw)
return save_image_with_bbox.counter
def save_image_with_bbox(image, labels_, scores_, bboxes_):
if not hasattr(save_image_with_bbox, "counter"):
save_image_with_bbox.counter = 0 # it doesn't exist yet, so initialize it
save_image_with_bbox.counter += 1
img_to_draw = np.copy(image)#common_preprocessing.np_image_unwhitened(image))
if _IN_DEBUG:
img_to_draw = draw_toolbox.bboxes_draw_on_img(img_to_draw, labels_, scores_, bboxes_, thickness=2)
imsave(os.path.join('./Debug/{}.jpg').format(save_image_with_bbox.counter), img_to_draw)
return save_image_with_bbox.counter#np.array([save_image_with_bbox.counter])
def save_image_with_bbox(image, labels_, scores_, bboxes_):
# 在图片上标记bounding boxes
if not hasattr(save_image_with_bbox, "counter"):
save_image_with_bbox.counter = 0 # 如果不存在，就初始化一个
save_image_with_bbox.counter += 1
img_to_draw = np.copy(image)
img_to_draw = draw_toolbox.bboxes_draw_on_img(img_to_draw,
labels_,
scores_,
bboxes_,
thickness=2) # 画框
imsave(
os.path.join('./debug/{}.jpg').format(save_image_with_bbox.counter),
img_to_draw)
return save_image_with_bbox.counter
def save_image_with_bbox(image, labels_, scores_, bboxes_):
# 存储带bounding boxes的图片
if not hasattr(save_image_with_bbox, "counter"):
save_image_with_bbox.counter = 0 # 如果不存在就初始化
save_image_with_bbox.counter += 1
img_to_draw = np.copy(image)
img_to_draw = draw_toolbox.bboxes_draw_on_img(img_to_draw,
labels_,
scores_,
bboxes_,
thickness=2)
imsave(
os.path.join(
'/home/yhq/Desktop/SSD-short/dataset/debug/{}.jpg').format(
save_image_with_bbox.counter), img_to_draw)
return save_image_with_bbox.counter
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2, ))
features = common_preprocessing.preprocess_for_test(
image_input,
out_shape,
data_format=('NCHW'
if FLAGS.data_format == 'channels_first' else 'NHWC'))
features = tf.expand_dims(features, axis=0)
anchor_creator = anchor_manipulator_v2.AnchorCreator(
out_shape,
layers_shapes=[(24, 24), (12, 12), (6, 6)],
anchor_scales=[(0.1, ), (0.2, 0.375, 0.55), (0.725, 0.9)],
extra_anchor_scales=[(0.1414, ), (0.2739, 0.4541, 0.6315),
(0.8078, 0.9836)],
anchor_ratios=[(2., .5), (2., 3., .5, 0.3333), (2., .5)],
layer_steps=[16, 32, 64])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors(
)
anchor_encoder_decoder = anchor_manipulator_v2.AnchorEncoder(
allowed_borders=[1.0] * 6,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
decode_fn = lambda pred: anchor_encoder_decoder.ext_decode_all_anchors(
pred, all_anchors, all_num_anchors_depth, all_num_anchors_spatial)
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
with tf.device('/gpu:0'):
backbone = xdet_body_v4.xdet_resnet_v4(FLAGS.resnet_size,
FLAGS.data_format)
backbone_outputs = backbone(inputs=features, is_training=False)
cls_pred, location_pred = xdet_body_v4.xdet_head(
backbone_outputs,
FLAGS.num_classes,
all_num_anchors_depth,
False,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1])
for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes])
for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
cls_pred = tf.concat(cls_pred, axis=0)
location_pred = tf.concat(location_pred, axis=0)
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0)
selected_bboxes, selected_scores = parse_by_class(
cls_pred, bboxes_pred, FLAGS.num_classes, FLAGS.select_threshold,
FLAGS.min_size, FLAGS.keep_topk, FLAGS.nms_topk,
FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
summary_dir = os.path.join(FLAGS.model_dir, 'predict')
if not os.path.exists(summary_dir):
os.makedirs(summary_dir)
all_images = tf.gfile.Glob(
os.path.join(FLAGS.test_dataset_path, '*.jpg'))
#print(all_images)
len_images = len(all_images)
config = tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False)
saver = tf.train.Saver()
with tf.Session(config=config) as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
for ind, image_name in enumerate(all_images):
sys.stdout.write('\r>> Processing image %d/%d' %
(ind + 1, len_images))
sys.stdout.flush()
np_image = imread(image_name)
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={
image_input: np_image,
shape_input: np_image.shape[:-1]
})
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
imsave(
os.path.join(FLAGS.debug_dir,
'output_{}.jpg'.format(image_name[-10:-4])),
img_to_draw)
sys.stdout.write('\n')
sys.stdout.flush()
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2, ))
features = ssd_preprocessing.preprocess_for_eval(
image_input,
out_shape,
data_format=FLAGS.data_format,
output_rgb=False)
features = tf.expand_dims(features, axis=0)
anchor_creator = anchor_manipulator.AnchorCreator(
out_shape,
layers_shapes=[(38, 38), (19, 19), (10, 10), (5, 5), (3, 3),
(1, 1)],
anchor_scales=[(0.1, ), (0.2, ), (0.375, ), (0.55, ), (0.725, ),
(0.9, )],
extra_anchor_scales=[(0.1414, ), (0.2739, ), (0.4541, ),
(0.6315, ), (0.8078, ), (0.9836, )],
anchor_ratios=[(2., .5),
(2., 3., .5, 0.3333), (2., 3., .5, 0.3333),
(2., 3., .5, 0.3333), (2., .5), (2., .5)],
layer_steps=[8, 16, 32, 64, 100, 300])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors(
)
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
allowed_borders=[1.0] * 6,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
decode_fn = lambda pred: anchor_encoder_decoder.ext_decode_all_anchors(
pred, all_anchors, all_num_anchors_depth, all_num_anchors_spatial)
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.forward(features, training=False)
location_pred, cls_pred = ssd_net.multibox_head(
feature_layers,
FLAGS.num_classes,
all_num_anchors_depth,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
cls_pred = tf.concat(cls_pred, axis=0)
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0)
selected_bboxes, selected_scores = parse_by_class(
cls_pred, bboxes_pred, FLAGS.num_classes,
FLAGS.select_threshold, FLAGS.min_size, FLAGS.keep_topk,
FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
np_image = imread('./demo/test.jpg')
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={
image_input: np_image,
shape_input: np_image.shape[:-1]
})
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
imsave('./demo/test_out.jpg', img_to_draw)
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
with tf.name_scope('define_input'):
image_input = tf.placeholder(tf.uint8,
shape=(None, None, 3),
name='image_input')
features = ssd_preprocessing.preprocess_for_eval(
image_input,
out_shape,
data_format=FLAGS.data_format,
output_rgb=False)
features = tf.expand_dims(features, axis=0)
anchor_creator = anchor_manipulator.AnchorCreator(
out_shape,
layers_shapes=[(38, 38), (19, 19), (10, 10), (5, 5), (3, 3),
(1, 1)],
anchor_scales=[(0.1, ), (0.2, ), (0.375, ), (0.55, ), (0.725, ),
(0.9, )],
extra_anchor_scales=[(0.1414, ), (0.2739, ), (0.4541, ),
(0.6315, ), (0.8078, ), (0.9836, )],
anchor_ratios=[(1., 2., .5), (1., 2., 3., .5, 0.3333),
(1., 2., 3., .5, 0.3333), (1., 2., 3., .5, 0.3333),
(1., 2., .5), (1., 2., .5)],
#anchor_ratios = [(2., .5), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333),
#(2., 3., .5, 0.3333), (2., .5), (2., .5)],
layer_steps=[8, 16, 32, 64, 100, 300])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors(
)
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
allowed_borders=[1.0] * 6,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
def decode_fn(pred):
return anchor_encoder_decoder.ext_decode_all_anchors(
pred, all_anchors, all_num_anchors_depth,
all_num_anchors_spatial)
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.forward(features, training=False)
location_pred, cls_pred = ssd_net.multibox_head(
feature_layers,
FLAGS.num_classes,
all_num_anchors_depth,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
with tf.variable_scope('cls_pred'):
cls_pred = tf.concat(cls_pred, axis=0)
with tf.variable_scope('location_pred'):
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0)
selected_bboxes, selected_scores = parse_by_class(
cls_pred, bboxes_pred, FLAGS.num_classes,
FLAGS.select_threshold, FLAGS.min_size, FLAGS.keep_topk,
FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
'''
config = tf.ConfigProto(allow_soft_placement=True, inter_op_parallelism_threads=1, intra_op_parallelism_threads=1)
config.mlu_options.data_parallelism = 1
config.mlu_options.model_parallelism = 1
config.mlu_options.core_num = 1
config.mlu_options.core_version = 'MLU270'
config.mlu_options.precision = 'float'
with tf.Session(config = config) as sess:
'''
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
np_image = imread('demo/test.jpg')
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={image_input: np_image})
#print('labels_', labels_, type(labels_), labels_.shape)
#print('scores_', scores_, type(scores_), scores_.shape)
#print('bboxes_', bboxes_, type(bboxes_), bboxes_.shape, bboxes_.shape[0])
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
imsave('demo/test_out.jpg', img_to_draw)
saver.save(sess, 'model/ssd300_vgg16/ssd300_vgg16', global_step=0)
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2, ))
features = ssd_preprocessing.preprocess_for_eval(
image_input,
out_shape,
data_format=FLAGS.data_format,
output_rgb=False)
features = tf.expand_dims(features, axis=0) #(N, W, H, C)
anchor_creator = anchor_manipulator.AnchorCreator(
out_shape,
layers_shapes=[(38, 38), (19, 19), (10, 10), (5, 5), (3, 3),
(1, 1)],
anchor_scales=[(0.1, ), (0.2, ), (0.375, ), (0.55, ), (0.725, ),
(0.9, )],
extra_anchor_scales=[(0.1414, ), (0.2739, ), (0.4541, ),
(0.6315, ), (0.8078, ), (0.9836, )],
anchor_ratios=[(1., 2., .5), (1., 2., 3., .5, 0.3333),
(1., 2., 3., .5, 0.3333), (1., 2., 3., .5, 0.3333),
(1., 2., .5), (1., 2., .5)],
#anchor_ratios = [(2., .5), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., .5), (2., .5)],
layer_steps=[8, 16, 32, 64, 100, 300])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors(
)
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
allowed_borders=[1.0] * 6,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
decode_fn = lambda pred: anchor_encoder_decoder.ext_decode_all_anchors(
pred, all_anchors, all_num_anchors_depth, all_num_anchors_spatial)
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
#backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
backbone = ssd_net.MobileNetV2Backbone(FLAGS.data_format)
feature_layers = backbone.forward(features, training=False)
location_pred, cls_pred = ssd_net.multibox_head(
feature_layers,
FLAGS.num_classes,
all_num_anchors_depth,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
cls_pred = tf.concat(cls_pred, axis=0)
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0) #
selected_bboxes, selected_scores = parse_by_class(
cls_pred, bboxes_pred, FLAGS.num_classes,
FLAGS.select_threshold, FLAGS.min_size, FLAGS.keep_topk,
FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
while (video.isOpened()):
ret, frame = video.read()
if ret == False:
break
else:
timer2 = cv2.getTickCount()
if undistort == 'y':
######################################## Undistortion Parts ########################################
dim2 = None
dim3 = None
timer = cv2.getTickCount()
dim1 = frame.shape[:
2][::
-1] # dim1 is the dimension of input image to un-distort
assert dim1[0] / dim1[1] == DIM[0] / DIM[
1], "Image to undistort needs to have same aspect ratio as the ones used in calibration"
if not dim2:
dim2 = dim1
if not dim3:
dim3 = dim1
scaled_K = K * dim1[0] / DIM[
0] # The values of K is to scale with image dimension.
scaled_K[2][
2] = 1.0 # Except that K[2][2] is always 1.0
# This is how scaled_K, dim2 and balance are used to determine the final K used to un-distort image. OpenCV document failed to make this clear!
new_K = cv2.fisheye.estimateNewCameraMatrixForUndistortRectify(
scaled_K, D, dim2, np.eye(3), balance=0)
map1, map2 = cv2.fisheye.initUndistortRectifyMap(
scaled_K, D, np.eye(3), new_K, dim3, cv2.CV_16SC2)
frame_r = cv2.remap(frame,
map1,
map2,
interpolation=cv2.INTER_LINEAR,
borderMode=cv2.BORDER_CONSTANT)
t = (cv2.getTickCount() -
timer) / cv2.getTickFrequency()
# frame_r = cv2.resize(dst, (640, 360))
# frame_r = cv2.putText(frame_r, "Undistortion processing time: %.3f sec" % t, (10, 40), cv2.FONT_HERSHEY_SIMPLEX, 0.5,(0, 255, 255), 2) #(1.0 / (end - start))
# ###############################################################################################
#np_image = imread('./demo/test.jpg')
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={
image_input: frame,
shape_input: frame.shape[:-1]
})
img_to_draw = draw_toolbox.bboxes_draw_on_img(frame,
labels_,
scores_,
bboxes_,
thickness=2)
fps = cv2.getTickFrequency() / (cv2.getTickCount() -
timer2)
img_to_draw = cv2.putText(img_to_draw, "FPS : %.1f" % fps,
(10, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5,
(0, 255, 255), 2) # dst_r
cv2.imshow('Object detector', img_to_draw) # dst_r
if cv2.waitKey(1) == ord('q'):
break
def ssd(path):
# def ssd_res(img_path):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2,))
features = ssd_preprocessing.preprocess_for_eval(image_input, out_shape, data_format=FLAGS.data_format, output_rgb=False)
features = tf.expand_dims(features, axis=0)
anchor_creator = anchor_manipulator.AnchorCrealog_device_placementtor(out_shape,
layers_shapes = [(38, 38), (19, 19), (10, 10), (5, 5), (3, 3), (1, 1)],
anchor_scales = [(0.1,), (0.2,), (0.375,), (0.55,), (0.725,), (0.9,)],
extra_anchor_scales = [(0.1414,), (0.2739,), (0.4541,), (0.6315,), (0.8078,), (0.9836,)],
anchor_ratios = [(1., 2., .5), (1., 2., 3., .5, 0.3333), (1., 2., 3., .5, 0.3333), (1., 2., 3., .5, 0.3333), (1., 2., .5), (1., 2., .5)],
#anchor_ratios = [(2., .5), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., .5), (2., .5)],
layer_steps = [8, 16, 32, 64, 100, 300])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors()
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(allowed_borders = [1.0] * 6,
positive_threshold = None,
ignore_threshold = None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
decode_fn = lambda pred : anchor_encoder_decoder.ext_decode_all_anchors(pred, all_anchors, all_num_anchors_depth, all_num_anchors_spatial)
with tf.variable_scope(FLAGS.model_scope, default_name=None, values=[features], reuse=tf.AUTO_REUSE):
backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.forward(features, training=False)
location_pred, cls_pred = ssd_net.multibox_head(feature_layers, FLAGS.num_classes, all_num_anchors_depth, data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred]
location_pred = [tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred]
cls_pred = [tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred]
location_pred = [tf.reshape(pred, [-1, 4]) for pred in location_pred]
cls_pred = tf.concat(cls_pred, axis=0)
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0)
selected_bboxes, selected_scores = parse_by_class(cls_pred, bboxes_pred,
FLAGS.num_classes, FLAGS.select_threshold, FLAGS.min_size,
FLAGS.keep_topk, FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
np_image = imread(path)
im = Image.open(path)
print(np_image.shape)
labels_, scores_, bboxes_ = sess.run([all_labels, all_scores, all_bboxes], feed_dict = {image_input : np_image, shape_input : np_image.shape[:-1]})
all_bboxes = sess.run([bboxes_pred], feed_dict = {image_input : np_image, shape_input : np_image.shape[:-1]})
shape = np_image.shape
for j in range(len(all_bboxes[0])):
all_box = all_bboxes[0][j]
p1 = (int(all_box[0] * shape[0]), int(all_box[1] * shape[1]))
p2 = (int(all_box[2] * shape[0]), int(all_box[3] * shape[1]))
if (p2[0] - p1[0] < 1) or (p2[1] - p1[1] < 1):
continue
x1 = p1[1]
y1 = p1[0]
x2 = p2[1]
y2 = p2[0]
obj = im.crop((x1, y1, x2, y2))
num_str = str(j)
num_str = num_str.zfill(5)
obj.save('./res/img/{}.jpg'.format(num_str))
cor = str(x1) + ',' + str(y1) + ',' + str(x2) + ',' +str(y2)
f2 = open('./res/cor.txt', 'a')
f2.write(cor + '\n')
zero_str = str(0)
f = open('./res/label.txt', 'a')
f.write(num_str + ',' + zero_str + '\n')
f.close()
num1 = 0
for i in range(bboxes_.shape[0]):
bbox = bboxes_[i]
p1 = (int(bbox[0] * shape[0]), int(bbox[1] * shape[1]))
p2 = (int(bbox[2] * shape[0]), int(bbox[3] * shape[1]))
num1 = num1 + 1
if (p2[0] - p1[0] < 1) or (p2[1] - p1[1] < 1):
continue
x1 = p1[1]
y1 = p1[0]
x2 = p2[1]
y2 = p2[0]
cor1 = str(x1) + ',' + str(y1) + ',' + str(x2) + ',' + str(y2)
num = 0
with open('./res/cor.txt', 'r') as f11, open('./res/label.txt', '+r') as f22:
for line in f11:
num = num + 1
if cor1 in line:
num11 = str(num)
print(num11 + '\n')
num11 = num11.zfill(5)
ber = num11 + ',' + str(0)
aft = num11 + ',' + str(labels_[i])
t = f22.read()
t = t.replace(ber, aft)
f22.seek(0, 0)
f22.write(t)
print(num1)
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image, labels_, scores_, bboxes_, thickness=2)
imsave('./demo/out.jpg', img_to_draw)
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
anchor_creator = anchor_manipulator.AnchorCreator(
out_shape,
layers_shapes=[(38, 38), (19, 19), (10, 10), (5, 5), (3, 3),
(1, 1)],
anchor_scales=[(0.1, ), (0.2, ), (0.375, ), (0.55, ), (0.725, ),
(0.9, )],
extra_anchor_scales=[(0.1414, ), (0.2739, ), (0.4541, ),
(0.6315, ), (0.8078, ), (0.9836, )],
anchor_ratios=[(1., 2., .5), (1., 2., 3., .5, 0.3333),
(1., 2., 3., .5, 0.3333), (1., 2., 3., .5, 0.3333),
(1., 2., .5), (1., 2., .5)],
#anchor_ratios = [(2., .5), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333),
#(2., 3., .5, 0.3333), (2., .5), (2., .5)],
layer_steps=[8, 16, 32, 64, 100, 300])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors(
)
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
allowed_borders=[1.0] * 6,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
def decode_fn(pred):
return anchor_encoder_decoder.ext_decode_all_anchors(
pred, all_anchors, all_num_anchors_depth,
all_num_anchors_spatial)
with tf.name_scope('define_input'):
image_input = tf.placeholder(tf.float32,
shape=(1, 300, 300, 3),
name='image_input')
print('image_input', image_input)
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[image_input],
reuse=tf.AUTO_REUSE):
backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.forward(image_input, training=False)
location_pred, cls_pred = ssd_net.multibox_head(
feature_layers,
FLAGS.num_classes,
all_num_anchors_depth,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
with tf.variable_scope('cls_pred'):
cls_pred = tf.concat(cls_pred, axis=0)
with tf.variable_scope('location_pred'):
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0)
selected_bboxes, selected_scores = parse_by_class(
cls_pred, bboxes_pred, FLAGS.num_classes,
FLAGS.select_threshold, FLAGS.min_size, FLAGS.keep_topk,
FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
'''
config = tf.ConfigProto(allow_soft_placement=True, inter_op_parallelism_threads=1, intra_op_parallelism_threads=1)
config.mlu_options.data_parallelism = 1
config.mlu_options.model_parallelism = 1
config.mlu_options.core_num = 1
config.mlu_options.core_version = 'MLU270'
config.mlu_options.precision = 'float'
with tf.Session(config = config) as sess:
'''
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
_R_MEAN = 123.68
_G_MEAN = 116.78
_B_MEAN = 103.94
means = [
_B_MEAN,
_G_MEAN,
_R_MEAN,
]
np_image = cv2.imread('demo/test.jpg')
image = cv2.resize(
np_image, (FLAGS.train_image_size, FLAGS.train_image_size))
image = (image - means) # / 255.0
image = np.expand_dims(image, axis=0)
print('image', type(image), image.shape)
'''
image = tf.to_float(np_image)
image = tf.image.resize_images(image, out_shape,
method=tf.image.ResizeMethod.BILINEAR, align_corners=False)
image.set_shape(out_shape + [3])
num_channels = image.get_shape().as_list()[-1]
channels = tf.split(axis=2, num_or_size_splits=num_channels, value=image)
for i in range(num_channels):
channels[i] -= means[i]
image = tf.concat(axis=2, values=channels)
image_channels = tf.unstack(image, axis=-1, name='split_rgb')
image = tf.stack([image_channels[2], image_channels[1], image_channels[0]], axis=-1, name='merge_bgr')
'''
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={image_input: image})
#print('labels_', labels_, type(labels_), labels_.shape)
#print('scores_', scores_, type(scores_), scores_.shape)
#print('bboxes_', bboxes_, type(bboxes_), bboxes_.shape, bboxes_.shape[0])
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
cv2.imwrite('demo/test_out.jpg', img_to_draw)
saver.save(sess,
'model/ssd300_vgg16/ssd300_vgg16_short',
global_step=0)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
print('sess2 start')
with tf.Session(graph=g2) as sess2:
print('sess2 end')
labels_, scores_, bboxes_ = sess2.run([all_labels, all_scores, all_bboxes],
feed_dict={
g2_cls_pred: cls_pred_,
g2_location_pred: location_pred_
})
#print('labels_', labels_, type(labels_), labels_.shape)
#print('scores_', scores_, type(scores_), scores_.shape)
#print('bboxes_', bboxes_, type(bboxes_), bboxes_.shape, bboxes_.shape[0])
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
imsave('demo/test_out.jpg', img_to_draw)
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2, ))
features, output_shape = ssd_preprocessing.preprocess_for_eval(
image_input,
out_shape,
data_format=FLAGS.data_format,
output_rgb=False)
features = tf.expand_dims(features, axis=0)
output_shape = tf.expand_dims(output_shape, axis=0)
all_anchor_scales = [(30., ), (60., ), (112.5, ), (165., ), (217.5, ),
(270., )]
all_extra_scales = [(42.43, ), (82.17, ), (136.23, ), (189.45, ),
(242.34, ), (295.08, )]
all_anchor_ratios = [(1., 2., .5), (1., 2., 3., .5, 0.3333),
(1., 2., 3., .5, 0.3333),
(1., 2., 3., .5, 0.3333), (1., 2., .5),
(1., 2., .5)]
# all_anchor_ratios = [(2., .5), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., .5), (2., .5)]
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.forward(features, training=False)
with tf.device('/cpu:0'):
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
if FLAGS.data_format == 'channels_first':
all_layer_shapes = [
tf.shape(feat)[2:] for feat in feature_layers
]
else:
all_layer_shapes = [
tf.shape(feat)[1:3] for feat in feature_layers
]
all_layer_strides = [8, 16, 32, 64, 100, 300]
total_layers = len(all_layer_shapes)
anchors_height = list()
anchors_width = list()
anchors_depth = list()
for ind in range(total_layers):
_anchors_height, _anchors_width, _anchor_depth = anchor_encoder_decoder.get_anchors_width_height(
all_anchor_scales[ind],
all_extra_scales[ind],
all_anchor_ratios[ind],
name='get_anchors_width_height{}'.format(ind))
anchors_height.append(_anchors_height)
anchors_width.append(_anchors_width)
anchors_depth.append(_anchor_depth)
anchors_ymin, anchors_xmin, anchors_ymax, anchors_xmax, _ = anchor_encoder_decoder.get_all_anchors(
tf.squeeze(output_shape, axis=0), anchors_height,
anchors_width, anchors_depth, [0.5] * total_layers,
all_layer_shapes, all_layer_strides, [0.] * total_layers,
[False] * total_layers)
location_pred, cls_pred = ssd_net.multibox_head(
feature_layers,
FLAGS.num_classes,
anchors_depth,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
cls_pred = tf.concat(cls_pred, axis=0)
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = anchor_encoder_decoder.decode_anchors(
location_pred, anchors_ymin, anchors_xmin, anchors_ymax,
anchors_xmax)
selected_bboxes, selected_scores = bbox_util.parse_by_class(
tf.squeeze(output_shape, axis=0), cls_pred, bboxes_pred,
FLAGS.num_classes, FLAGS.select_threshold, FLAGS.min_size,
FLAGS.keep_topk, FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
np_image = imread('./demo/test.jpg')
labels_, scores_, bboxes_, output_shape_ = sess.run(
[all_labels, all_scores, all_bboxes, output_shape],
feed_dict={
image_input: np_image,
shape_input: np_image.shape[:-1]
})
bboxes_[:,
0] = bboxes_[:, 0] * np_image.shape[0] / output_shape_[0,
0]
bboxes_[:,
1] = bboxes_[:, 1] * np_image.shape[1] / output_shape_[0,
1]
bboxes_[:,
2] = bboxes_[:, 2] * np_image.shape[0] / output_shape_[0,
0]
bboxes_[:,
3] = bboxes_[:, 3] * np_image.shape[1] / output_shape_[0,
1]
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
imsave('./demo/test_out.jpg', img_to_draw)
def main(_):
with tf.Graph().as_default():
target_shape = None
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
features, output_shape = sfd_preprocessing.preprocess_for_eval(image_input, target_shape, data_format=FLAGS.data_format, output_rgb=False)
features = tf.expand_dims(features, axis=0)
output_shape = tf.expand_dims(output_shape, axis=0)
all_anchor_scales = [(16.,), (32.,), (64.,), (128.,), (256.,), (512.,)]
all_extra_scales = [(), (), (), (), (), ()]
all_anchor_ratios = [(1.,), (1.,), (1.,), (1.,), (1.,), (1.,)]
all_layer_strides = [4, 8, 16, 32, 64, 128]
offset_list = [0.5, 0.5, 0.5, 0.5, 0.5, 0.5]
with tf.variable_scope(FLAGS.model_scope, default_name=None, values=[features], reuse=tf.AUTO_REUSE):
backbone = sfd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.get_featmaps(features, training=False)
with tf.device('/cpu:0'):
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(positive_threshold=None, ignore_threshold=None, prior_scaling=[0.1, 0.1, 0.2, 0.2])
if FLAGS.data_format == 'channels_first':
all_layer_shapes = [tf.shape(feat)[2:] for feat in feature_layers]
else:
all_layer_shapes = [tf.shape(feat)[1:3] for feat in feature_layers]
total_layers = len(all_layer_shapes)
anchors_height = list()
anchors_width = list()
anchors_depth = list()
for ind in range(total_layers):
_anchors_height, _anchors_width, _anchor_depth = anchor_encoder_decoder.get_anchors_width_height(all_anchor_scales[ind], all_extra_scales[ind], all_anchor_ratios[ind], name='get_anchors_width_height{}'.format(ind))
anchors_height.append(_anchors_height)
anchors_width.append(_anchors_width)
anchors_depth.append(_anchor_depth)
anchors_ymin, anchors_xmin, anchors_ymax, anchors_xmax, _ = anchor_encoder_decoder.get_all_anchors(tf.squeeze(output_shape, axis=0),
anchors_height, anchors_width, anchors_depth,
offset_list, all_layer_shapes, all_layer_strides,
[0.] * total_layers, [False] * total_layers)
location_pred, cls_pred = backbone.multibox_head(feature_layers, [1] * len(feature_layers),
[3] + [1] * (len(feature_layers) - 1), anchors_depth)
if FLAGS.data_format == 'channels_first':
cls_pred = [tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred]
location_pred = [tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred]
cls_pred = [tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred]
location_pred = [tf.reshape(pred, [-1, 4]) for pred in location_pred]
cls_pred = tf.nn.softmax(tf.concat(cls_pred, axis=0))[:, -1]
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = anchor_encoder_decoder.decode_anchors(location_pred, anchors_ymin, anchors_xmin, anchors_ymax, anchors_xmax)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
os.makedirs(FLAGS.det_dir, exist_ok=True)
if FLAGS.subset is 'val':
wider_face = sio.loadmat(os.path.join(FLAGS.data_dir, 'wider_face_split', 'wider_face_val.mat')) # Val set
else:
wider_face = sio.loadmat(os.path.join(FLAGS.data_dir, 'wider_face_split', 'wider_face_test.mat')) # Test set
event_list = wider_face['event_list']
file_list = wider_face['file_list']
del wider_face
Path = os.path.join(FLAGS.data_dir, ('WIDER_val' if FLAGS.subset is 'val' else 'WIDER_test'), 'images')
save_path = os.path.join(FLAGS.det_dir, FLAGS.subset)
len_event = len(event_list)
for index, event in enumerate(event_list):
filelist = file_list[index][0]
len_files = len(filelist)
if not os.path.exists(os.path.join(save_path, event[0][0])):
os.makedirs(os.path.join(save_path, event[0][0]))
for num, file in enumerate(filelist):
im_name = file[0][0]
Image_Path = os.path.join(Path, event[0][0], im_name[:]+'.jpg')
image = imread(Image_Path)
#image = imread('manymany.jpg')
max_im_shrink = (0x7fffffff / FLAGS.memory_limit / (image.shape[0] * image.shape[1])) ** 0.5 # the max size of input image for caffe
#max_im_shrink = (0x7fffffff / 80.0 / (image.shape[0] * image.shape[1])) ** 0.5 # the max size of input image for caffe
shrink = max_im_shrink if max_im_shrink < 1 else 1
det0 = detect_face([sess, image_input, bboxes_pred, cls_pred], image, shrink) # origin test
det1 = flip_test([sess, image_input, bboxes_pred, cls_pred], image, shrink) # flip test
[det2, det3] = multi_scale_test([sess, image_input, bboxes_pred, cls_pred], image, max_im_shrink) #multi-scale test
# merge all test results via bounding box voting
det = np.row_stack((det0, det1, det2, det3))
dets = bbox_vote(det)
f = open(os.path.join(save_path, event[0][0], im_name+'.txt'), 'w')
write_to_txt(f, dets, event, im_name)
f.close()
if num % FLAGS.log_every_n_steps == 0:
img_to_draw = draw_toolbox.bboxes_draw_on_img(image, (dets[:, 4] > 0.2).astype(np.int32), dets[:, 4], dets[:, :4], thickness=2)
imsave(os.path.join(FLAGS.debug_dir, '{}.jpg'.format(im_name)), img_to_draw)
#imsave(os.path.join('./debug/{}_{}.jpg').format(index, num), draw_toolbox.absolute_bboxes_draw_on_img(image, (dets[:, 4]>0.1).astype(np.int32), dets[:, 4], dets[:, :4], thickness=2))
#break
sys.stdout.write('\r>> Predicting event:%d/%d num:%d/%d' % (index + 1, len_event, num + 1, len_files))
sys.stdout.flush()
sys.stdout.write('\n')
sys.stdout.flush()
def main(_):
with tf.Graph().as_default():
out_shape = [FLAGS.train_image_size] * 2
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2, ))
features = ssd_preprocessing.preprocess_for_eval(
image_input,
out_shape,
data_format=FLAGS.data_format,
output_rgb=False)
features = tf.expand_dims(features, axis=0)
anchor_creator = anchor_manipulator.AnchorCreator(
out_shape,
layers_shapes=[(38, 38), (19, 19), (10, 10), (5, 5), (3, 3),
(1, 1)],
anchor_scales=[(0.1, ), (0.2, ), (0.375, ), (0.55, ), (0.725, ),
(0.9, )],
extra_anchor_scales=[(0.1414, ), (0.2739, ), (0.4541, ),
(0.6315, ), (0.8078, ), (0.9836, )],
anchor_ratios=[(1., 2., .5), (1., 2., 3., .5, 0.3333),
(1., 2., 3., .5, 0.3333), (1., 2., 3., .5, 0.3333),
(1., 2., .5), (1., 2., .5)],
#anchor_ratios = [(2., .5), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., 3., .5, 0.3333), (2., .5), (2., .5)],
layer_steps=[8, 16, 32, 64, 100, 300])
all_anchors, all_num_anchors_depth, all_num_anchors_spatial = anchor_creator.get_all_anchors(
)
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
allowed_borders=[1.0] * 6,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[0.1, 0.1, 0.2, 0.2])
decode_fn = lambda pred: anchor_encoder_decoder.ext_decode_all_anchors(
pred, all_anchors, all_num_anchors_depth, all_num_anchors_spatial)
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
backbone = ssd_net.VGG16Backbone(FLAGS.data_format)
feature_layers = backbone.forward(features, training=False)
location_pred, cls_pred = ssd_net.multibox_head(
feature_layers,
FLAGS.num_classes,
all_num_anchors_depth,
data_format=FLAGS.data_format)
if FLAGS.data_format == 'channels_first':
cls_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in cls_pred
]
location_pred = [
tf.transpose(pred, [0, 2, 3, 1]) for pred in location_pred
]
cls_pred = [
tf.reshape(pred, [-1, FLAGS.num_classes]) for pred in cls_pred
]
location_pred = [
tf.reshape(pred, [-1, 4]) for pred in location_pred
]
cls_pred = tf.concat(cls_pred, axis=0)
location_pred = tf.concat(location_pred, axis=0)
with tf.device('/cpu:0'):
bboxes_pred = decode_fn(location_pred)
bboxes_pred = tf.concat(bboxes_pred, axis=0)
selected_bboxes, selected_scores = parse_by_class(
cls_pred, bboxes_pred, FLAGS.num_classes,
FLAGS.select_threshold, FLAGS.min_size, FLAGS.keep_topk,
FLAGS.nms_topk, FLAGS.nms_threshold)
labels_list = []
scores_list = []
bboxes_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
scores_list.append(v)
bboxes_list.append(selected_bboxes[k])
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(scores_list, axis=0)
all_bboxes = tf.concat(bboxes_list, axis=0)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, get_checkpoint())
for i in range(video_frame_cnt):
ret, img_ori = vid.read()
# height_ori, width_ori = img_ori.shape[:2]
# img = cv2.resize(img_ori, tuple(args.new_size))
img = cv2.cvtColor(img_ori, cv2.COLOR_BGR2RGB)
np_image = np.asarray(img, np.float32)
start_time = time.time()
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={
image_input: np_image,
shape_input: np_image.shape[:-1]
})
end_time = time.time()
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
cv2.putText(img_to_draw,
'{:.2f}ms'.format((end_time - start_time) * 1000),
(40, 40),
0,
fontScale=1,
color=(0, 255, 0),
thickness=2)
imsave('./test_out.jpg', img_to_draw)
new_img = cv2.imread('./test_out.jpg')
cv2.imshow('image', new_img)
videoWriter.write(new_img)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
vid.release()
videoWriter.release()
def main(_):
with tf.Graph().as_default():
image_input = tf.placeholder(tf.uint8, shape=(None, None, 3))
shape_input = tf.placeholder(tf.int32, shape=(2, ))
features = common_preprocessing.light_head_preprocess_for_test(
image_input, [FLAGS.train_image_size] * 2,
data_format=('NCHW'
if FLAGS.data_format == 'channels_first' else 'NHWC'))
features = tf.expand_dims(features, axis=0)
anchor_creator = anchor_manipulator.AnchorCreator(
[FLAGS.train_image_size] * 2,
layers_shapes=[(30, 30)],
anchor_scales=[[0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8]],
extra_anchor_scales=[[0.1]],
anchor_ratios=[[1., 2., .5]],
layer_steps=[16])
all_anchors, num_anchors_list = anchor_creator.get_all_anchors()
anchor_encoder_decoder = anchor_manipulator.AnchorEncoder(
all_anchors,
num_classes=FLAGS.num_classes,
allowed_borders=None,
positive_threshold=None,
ignore_threshold=None,
prior_scaling=[1., 1., 1., 1.])
with tf.variable_scope(FLAGS.model_scope,
default_name=None,
values=[features],
reuse=tf.AUTO_REUSE):
rpn_feat_map, backbone_feat = xception_body.XceptionBody(
features,
FLAGS.num_classes,
is_training=False,
data_format=FLAGS.data_format)
#rpn_feat_map = tf.Print(rpn_feat_map,[tf.shape(rpn_feat_map), rpn_feat_map,backbone_feat])
rpn_cls_score, rpn_bbox_pred = xception_body.get_rpn(
rpn_feat_map, num_anchors_list[0], False, FLAGS.data_format,
'rpn_head')
large_sep_feature = xception_body.large_sep_kernel(
backbone_feat, 256, 10 * 7 * 7, False, FLAGS.data_format,
'large_sep_feature')
if FLAGS.data_format == 'channels_first':
rpn_cls_score = tf.transpose(rpn_cls_score, [0, 2, 3, 1])
rpn_bbox_pred = tf.transpose(rpn_bbox_pred, [0, 2, 3, 1])
rpn_cls_score = tf.reshape(rpn_cls_score, [-1, 2])
rpn_object_score = tf.nn.softmax(rpn_cls_score)[:, -1]
rpn_object_score = tf.reshape(rpn_object_score, [1, -1])
rpn_location_pred = tf.reshape(rpn_bbox_pred, [1, -1, 4])
rpn_bboxes_pred = anchor_encoder_decoder.decode_all_anchors(
[rpn_location_pred], squeeze_inner=True)[0]
proposals_bboxes = xception_body.get_proposals(
rpn_object_score, rpn_bboxes_pred, None,
FLAGS.rpn_pre_nms_top_n, FLAGS.rpn_post_nms_top_n,
FLAGS.rpn_nms_thres, FLAGS.rpn_min_size, False,
FLAGS.data_format)
cls_score, bboxes_reg = xception_body.get_head(
large_sep_feature, lambda input_, bboxes_,
grid_width_, grid_height_: ps_roi_align(
input_, bboxes_, grid_width_, grid_height_, pool_method),
7, 7, None, proposals_bboxes, FLAGS.num_classes, False, False,
0, FLAGS.data_format, 'final_head')
head_bboxes_pred = anchor_encoder_decoder.ext_decode_rois(
proposals_bboxes,
bboxes_reg,
head_prior_scaling=[1., 1., 1., 1.])
head_cls_score = tf.reshape(cls_score, [-1, FLAGS.num_classes])
head_cls_score = tf.nn.softmax(head_cls_score)
head_bboxes_pred = tf.reshape(head_bboxes_pred, [-1, 4])
with tf.device('/device:CPU:0'):
selected_scores, selected_bboxes = eval_helper.tf_bboxes_select(
[head_cls_score], [head_bboxes_pred],
FLAGS.select_threshold,
FLAGS.num_classes,
scope='xdet_v2_select')
selected_bboxes = eval_helper.bboxes_clip(
tf.constant([0., 0., 1., 1.]), selected_bboxes)
selected_scores, selected_bboxes = eval_helper.filter_boxes(
selected_scores,
selected_bboxes,
0.03,
shape_input, [FLAGS.train_image_size] * 2,
keep_top_k=FLAGS.nms_topk * 2)
# Resize bboxes to original image shape.
selected_bboxes = eval_helper.bboxes_resize(
tf.constant([0., 0., 1., 1.]), selected_bboxes)
selected_scores, selected_bboxes = eval_helper.bboxes_sort(
selected_scores, selected_bboxes, top_k=FLAGS.nms_topk * 2)
# Apply NMS algorithm.
selected_scores, selected_bboxes = eval_helper.bboxes_nms_batch(
selected_scores,
selected_bboxes,
nms_threshold=FLAGS.nms_threshold,
keep_top_k=FLAGS.nms_topk)
labels_list = []
for k, v in selected_scores.items():
labels_list.append(tf.ones_like(v, tf.int32) * k)
all_labels = tf.concat(labels_list, axis=0)
all_scores = tf.concat(list(selected_scores.values()), axis=0)
all_bboxes = tf.concat(list(selected_bboxes.values()), axis=0)
saver = tf.train.Saver()
with tf.Session() as sess:
init = tf.global_variables_initializer()
sess.run(init)
saver.restore(sess, FLAGS.checkpoint_path)
np_image = imread('./demo/test.jpg')
labels_, scores_, bboxes_ = sess.run(
[all_labels, all_scores, all_bboxes],
feed_dict={
image_input: np_image,
shape_input: np_image.shape[:-1]
})
img_to_draw = draw_toolbox.bboxes_draw_on_img(np_image,
labels_,
scores_,
bboxes_,
thickness=2)
imsave(os.path.join(FLAGS.debug_dir, 'test_out.jpg'), img_to_draw)
