def callback(self, image_msg):
frame = self.bridge.imgmsg_to_cv2(image_msg, "bgr8")
image_data = cv2.imencode('.jpg', frame)[1].tostring()
image_data2 = cv2.imencode('.jpg', frame)[1]
image_data3 = cv2.imdecode(image_data2, cv2.IMREAD_COLOR)
# cv2.imshow("imgNet_tf", image_data3)
# cv2.waitKey(1)
cv2.putText(
image_data3,
self.human_string + ": " + str(int(self.score * 10000)) + "%",
(10, 30), 0, 1, (0, 8, 255), 2)
msg = self.bridge.cv2_to_imgmsg(image_data3, "bgr8")
self._pub1.publish(msg)
rospy.sleep(0.9)
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)
# Creates node ID --> English string lookup.
node_lookup = classify_image.NodeLookup()
top_k = predictions.argsort()[-self.use_top_k:][::-1]
for node_id in top_k:
self.human_string = node_lookup.id_to_string(node_id)
self.score = predictions[node_id]
if self.score > self.score_threshold:
rospy.logdebug('%s (score = %.5f)' %
(self.human_string, self.score))
self._pub.publish(self.human_string)
#topic_diag_main("ros_tf_node_imgNet", "", human_string, (float(score)) * 100)
data = {}
data['class'] = self.human_string
data['probablity'] = (float(self.score)) * 100
self.monitor.node_info(data, 'Result',
'Detection result received.', 0)
data3 = {}
data3['Status'] = 1
self.monitor.node_info(data3, 'Status',
'Detection status received.', 0)
def callback(self, image_msg):
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')
predicitons = self._session.run( softmax_tensor, {'DecodeJpeg/contents:0': image_data})
node_lookup = classify_image.NodeLookup()
top_k = predictions.argsort()[-self.use_top_k:][::-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):
cv_image = self._cv_bridge.imgmsg_to_cv2(image_msg, "bgr8")
# copy from
# 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')
predictions = self._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()[-self.use_top_k:][::-1]
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
#print(score, human_string)
if score > self.score_threshold:
rospy.loginfo('%s (score = %.5f)' % (human_string, score))
self._pub.publish(human_string)
def callback(self, image_msg):
# 将ROS的图像数据转换成OpenCV的图像格式
cv_image = self._cv_bridge.imgmsg_to_cv2(image_msg, "bgr8")
# copy from
# classify_image.py
image_data = cv2.imencode('.jpg', cv_image)[1].tostring()
# 从保存的GraphDef中创建softmax张量，其中softmax:0部分包含1000个标签化预测的张量
softmax_tensor = self._session.graph.get_tensor_by_name('softmax:0')
# 运行softmax张量，其中DecodeJpeg/contents:0行是一个包含图像JPEG编码字符串的张量
predictions = self._session.run(
softmax_tensor, {'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)
# 寻找匹配对象的字符串及其概率
node_lookup = classify_image.NodeLookup()
top_k = predictions.argsort()[-self.use_top_k:][::-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 run(self):
sqs = session.resource('sqs', 'us-west-1')
s3 = session.resource('s3')
request_queue = sqs.get_queue_by_name(QueueName=SQS_REQUEST_QUEUE_NAME)
response_queue = sqs.get_queue_by_name(
QueueName=SQS_RESPONSE_QUEUE_NAME)
terminate_queue = sqs.get_queue_by_name(
QueueName=SQS_TERMINATE_QUEUE_NAME)
bucket = s3.Bucket(S3_BUCKET_NAME)
# Creates node ID --> English string lookup.
node_lookup = classify_image.NodeLookup()
request_count = 0
with tf.Session() as sess:
softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
check = False
while True:
if check:
check = False
request_count = 0
while True:
for message in terminate_queue.receive_messages(
MaxNumberOfMessages=1,
AttributeNames=['SentTimestamp']):
message.delete()
message_ts = int(
message.attributes['SentTimestamp'])
current_ts = int(time.time() * 1000)
print current_ts, message_ts
if (current_ts - message_ts
) <= TERMINATE_REQUEST_EXPIRES_IN:
# Valid termination request
return
else:
# Expired request
break
else:
break
processed = False
for message in request_queue.receive_messages(
WaitTimeSeconds=20, MaxNumberOfMessages=1):
request = json.loads(message.body)
message.delete()
image_url = request['imageUrl']
image_data = self.download_image(image_url)
predictions = sess.run(
softmax_tensor, {'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)
node_id = predictions.argsort()[-1]
human_string = node_lookup.id_to_string(node_id)
print image_url, human_string
request['result'] = human_string
response_queue.send_message(
MessageBody=json.dumps(request))
if S3_UPLOAD_ENABLED:
image_name = os.path.basename(image_url)
while True:
try:
bucket.put_object(Key=image_name,
Body=human_string)
break
except Exception:
bucket = s3.Bucket(S3_BUCKET_NAME)
time.sleep(2)
processed = True
request_count += 1
if not processed:
check = True
elif request_count and (request_count %
TERMINATE_CHECK_INTERVAL == 0):
check = True
from flask import Flask, render_template, request
import classify_image
import tensorflow as tf
# setup flags
classify_image.FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string('model_dir', '/tmp/imagenet',
"""Path to classify_image_graph_def.pb""")
# prepare to using trained model
classify_image.maybe_download_and_extract()
classify_image.create_graph()
node_lookup = classify_image.NodeLookup()
sess = tf.Session()
softmax_tensor = tf.squeeze(sess.graph.get_tensor_by_name('softmax:0'))
app = Flask(__name__)
@app.route('/recognize', methods=['POST'])
def recognize():
f = request.files['image']
predictions = sess.run(softmax_tensor,
feed_dict={'DecodeJpeg/contents:0': f.read()})
results = []
top_k = predictions.argsort()[-5:][::-1]
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
results.append({'label': human_string, 'score': score})
return render_template('result.html', results=results)