def image_recognition(cv_image):
#cv2.imshow('image',cv_image)
# cv2.waitKey(1000)
image_data = cv2.imencode('.jpg', cv_image)[1].tostring()
# Creates graph from saved GraphDef.
softmax_tensor = session.graph.get_tensor_by_name('softmax:0')
predictions = session.run(softmax_tensor,
{'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)
# Creates node ID --> English string lookup.
node_lookup = classify_image.NodeLookup()
top_k = predictions.argsort()[use_top_k:][::-1]
strings = []
scores = []
result_num = 0
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
if score > score_threshold:
strings.append(human_string)
scores.append(float(score))
result_num += 1
return strings, scores, result_num
def callback(self, image_msg):
cv_image = self._cv_bridge.imgmsg_to_cv2(image_msg, "bgr8")
# copy from
# https://github.com/tensorflow/tensorflow/blob/master/tensorflow/models/image/imagenet/classify_image.py
image_data = cv2.imencode('.jpg', cv_image)[1].tostring()
# Creates graph from saved GraphDef.
softmax_tensor = self._session.graph.get_tensor_by_name('softmax:0')
# Creates node ID --> English string lookup.
node_lookup = classify_image.NodeLookup()
top_k = predictions.argsort(
)[-self.use_top_number_of_predictions:][::-1]
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
if score > self.score_threshold:
rospy.loginfo('%s (score = %.5f)' % (human_string, score))
self._pub.publish(human_string)
def callback(self, image_msg):
# convert image from ROS image msg type to opencv format
cv_image = self._cv_bridge.imgmsg_to_cv2(image_msg, "bgr8")
image_data = cv2.imencode('.jpg', cv_image)[1].tostring()
softmax_tensor = self._session.graph.get_tensor_by_name('softmax:0')
predictions = self._session.run(
softmax_tensor, {'DecodeJpeg/contents:0':image_data}
)
predictions = np.squeeze(predictions)
# create nodeID -- english string lookup
node_lookup = classify_image.NodeLookup()
top_k = predictions.argsort()[-self.use_top_k:][::-1] # get top k
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
if score > self.score_threshold:
rospy.loginfo('%s (score = %.5f)' %(human_string, score))
self._pub.publish(human_string)
def run_top_k_predictions_on_image(image):
if not tf.gfile.Exists(image):
tf.logging.fatal('File does not exist %s', image)
image_data = tf.gfile.FastGFile(image, 'rb').read()
# Download the incepion-v3 model
ic.maybe_download_and_extract()
# Creates graph from saved GraphDef.
ic.create_graph()
with tf.Session() as sess:
# Convert the PNG image to a height x width x 3 (channels) Numpy array,
# for example using PIL, then feed the 'DecodeJpeg:0' tensor:
image_content = Image.open(image)
image_array = np.array(image_content)[:, :, 0:3] # Select RGB channels only.
# Some useful tensors:
# 'softmax:0': A tensor containing the normalized prediction across
# 1000 labels.
# 'pool_3:0': A tensor containing the next-to-last layer containing 2048
# float description of the image.
# 'DecodeJpeg/contents:0': A tensor containing a string providing JPEG
# encoding of the image.
# Runs the softmax tensor by feeding the image_data as input to the graph.
softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
predictions = sess.run(softmax_tensor, {'DecodeJpeg:0': image_array})
predictions = np.squeeze(predictions)
# Creates node ID --> English string lookup.
node_lookup = ic.NodeLookup()
top_k = predictions.argsort()[-ic.FLAGS.num_top_predictions:][::-1]
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
print('%s (score = %.5f)' % (human_string, score))
def _classify_cb(self, image):
try:
cv_img = self._BRIDGE.imgmsg_to_cv2(image)
except CvBridgeError as e:
rospy.logerr("cv_bridge exception: {}".format(e))
img_data = cv2.imencode('.jpg', cv_img)[1].tostring()
softmax = self._sess.graph.get_tensor_by_name('softmax:0')
pred = self._sess.run(softmax, {'DecodeJpeg/contents:0': img_data})
pred = np.squeeze(pred)
lookup = classify_image.NodeLookup()
top_k = pred.argsort()[-self._k:][::-1]
ss = []
for id in top_k:
ret_str = lookup.id_to_string(id)
score = pred[id]
if score > self._score_threshold:
ss.append("[{:0.4f} - {}]\n".format(score, ret_str))
self._string_pub.publish(','.join(ss))
self._show_image(cv_img, ss[0][:-1])