def detect(self, img):
'''
:param img:
:return: final result:
'''
img, scale = self.resize_im(img,
scale=TextLineCfg.SCALE,
max_scale=TextLineCfg.MAX_SCALE)
blobs, im_scales = _get_blobs(img, None)
if cfg.TEST.HAS_RPN:
im_blob = blobs['data']
blobs['im_info'] = np.array(
[[im_blob.shape[1], im_blob.shape[2], im_scales[0]]],
dtype=np.float32)
cls_prob, box_pred = self.sess.run(
[self.output_cls_prob, self.output_box_pred],
feed_dict={self.input_img: blobs['data']})
rois, _ = proposal_layer(cls_prob,
box_pred,
blobs['im_info'],
'TEST',
anchor_scales=cfg.ANCHOR_SCALES)
scores = rois[:, 0]
boxes = rois[:, 1:5] / im_scales[0]
textdetector = TextDetector()
boxes = textdetector.detect(boxes, scores[:, np.newaxis],
img.shape[:2])
return self.draw_boxes(img, boxes, scale)
def ctpn(img):
timer = Timer()
timer.tic()
img, scale = resize_im(img, scale=TextLineCfg.SCALE, max_scale=TextLineCfg.MAX_SCALE)
scores, boxes = test_ctpn(sess, net, img)
textdetector = TextDetector()
boxes = textdetector.detect(boxes, scores[:, np.newaxis], img.shape[:2])
timer.toc()
print("\n----------------------------------------------")
print(('Detection took {:.3f}s for '
'{:d} object proposals').format(timer.total_time, boxes.shape[0]))
return scores, boxes, img, scale
def ctpn(self, image_name):
timer = Timer()
timer.tic()
img = cv2.imread(image_name)
img, scale = self.resize_im(img,
scale=TextLineCfg.SCALE,
max_scale=TextLineCfg.MAX_SCALE)
scores, boxes = test_ctpn(self.sess, self.net, img)
textdetector = TextDetector()
boxes = textdetector.detect(boxes, scores[:, np.newaxis],
img.shape[:2])
timer.toc()
print(('Detection took {:.3f}s for '
'{:d} object proposals').format(timer.total_time,
boxes.shape[0]))
return self.draw_boxes(img, image_name, boxes, scale)
def predict():
clearData()
cfg_from_file(r'./ctpn/ctpn/text.yml')
# init session
config = tf.ConfigProto(allow_soft_placement=True)
config.gpu_options.per_process_gpu_memory_fraction = 0.4
sess = tf.Session(config=config)
with gfile.FastGFile(r'.\ctpn\ctpn\data\ctpn.pb', 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
sess.graph.as_default()
tf.import_graph_def(graph_def, name='')
sess.run(tf.global_variables_initializer())
input_img = sess.graph.get_tensor_by_name('Placeholder:0')
output_cls_prob = sess.graph.get_tensor_by_name('Reshape_2:0')
output_box_pred = sess.graph.get_tensor_by_name(
'rpn_bbox_pred/Reshape_1:0')
im_names = os.listdir('./demo/test_images')
index = 0
for im_name in im_names:
print('~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~')
print(('finding card number for {:s}'.format(im_name)))
img = cv2.imread('./demo/test_images' + '//' + im_name)
img, scale = resize_im(img,
scale=TextLineCfg.SCALE,
max_scale=TextLineCfg.MAX_SCALE)
blobs, im_scales = _get_blobs(img, None)
if cfg.TEST.HAS_RPN:
im_blob = blobs['data']
blobs['im_info'] = np.array(
[[im_blob.shape[1], im_blob.shape[2], im_scales[0]]],
dtype=np.float32)
cls_prob, box_pred = sess.run([output_cls_prob, output_box_pred],
feed_dict={input_img: blobs['data']})
rois, _ = proposal_layer(cls_prob,
box_pred,
blobs['im_info'],
'TEST',
anchor_scales=cfg.ANCHOR_SCALES)
scores = rois[:, 0]
boxes = rois[:, 1:5] / im_scales[0]
textdetector = TextDetector()
boxes = textdetector.detect(boxes, scores[:, np.newaxis],
img.shape[:2])
draw_boxes(img, im_name, boxes, scale, index)
index += 1
print('recognizing card number:')
crnn = CRNN(image_shape=crnn_config.image_shape,
min_len=crnn_config.min_len,
max_len=crnn_config.max_len,
lstm_hidden=crnn_config.lstm_hidden,
pool_size=crnn_config.pool_size,
learning_decay_rate=crnn_config.learning_decay_rate,
learning_rate=crnn_config.learning_rate,
learning_decay_steps=crnn_config.learning_decay_steps,
mode=crnn_config.mode,
dict=crnn_config.dict,
is_training=True,
train_label_path=crnn_config.predict_label_path,
train_images_path=crnn_config.predict_images_path,
charset_path=crnn_config.charset_path)
result = crnn.predict(epoch=crnn_config.epoch,
batch_size=crnn_config.batch_size,
train_images_path=crnn_config.cardNum_path,
train_label_path=crnn_config.predict_label_path,
restore=True,
fonts=crnn_config.fonts,
logs_path=crnn_config.logs_path,
models_path=crnn_config.models_path)
return result
time5 = time.time()
for im_name in im_names:
print('~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~')
print(('Demo for {:s}'.format(im_name)))
img = cv2.imread(im_name)
img, scale = resize_im(img, scale=600, max_scale=1200) ##缩放至合适比例##
blobs, im_scales = _get_blobs(img, None) ##得到输入图片和缩放系数##
if True:
im_blob = blobs['data']
blobs['im_info'] = np.array(
[[im_blob.shape[1], im_blob.shape[2], im_scales[0]]],
dtype=np.float32)
cls_prob, box_pred = sess.run([output_cls_prob, output_box_pred], feed_dict={input_img: blobs['data']})
rois, _ = proposal_layer(cls_prob, box_pred, blobs['im_info'], 'TEST', anchor_scales=cfg.ANCHOR_SCALES)
scores = rois[:, 0]
boxes = rois[:, 1:5] / im_scales[0]
textdetector = TextDetector()
boxes = textdetector.detect(boxes, scores[:, np.newaxis], img.shape[:2])
generate_detect_box(img, im_name, boxes, scale)
#sort_box()
time6 = time.time()
print("text detect", time6 - time5)
time7 = time.time()