def predict(image, sess, softmax_tensor):
"""
Function used by classification workers to get prediction on image.
This method was adapted based on run_inference_on_image() method from classify_image.py found in TensorFlow official tutorial.
:param image: filename of the image to be classified
:param sess: TensorFlow session
:param softmax_tensor: tensor used for computing the predictions
:return: (img_id, result) with img_id being Instagram Image ID and the result being dictionary with 5 most probable
objects depicted in the image as keys and corresponding prediction confidences as values.
"""
img_id = os.path.splitext(os.path.basename(image))[0]
image_data = tf.gfile.FastGFile(image, 'rb').read()
# 'DecodeJpeg/contents:0': A tensor containing a string providing JPEG
# encoding of the image.
predictions = sess.run(softmax_tensor,
{'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)
# Creates node ID --> English string lookup.
node_lookup = NodeLookup()
top_k = predictions.argsort()[-FLAGS.num_top_predictions:][::-1]
result = {}
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
result[human_string] = float(score)
return img_id, result
def run_inference_on_image(self, imageList, num_top_predictions=5):
"""Runs inference on an image.
Args:
image: Image file name.
Returns:
Nothing
"""
for image in imageList:
if not tf.gfile.Exists(image):
tf.logging.fatal('File does not exist %s', image)
image_data = tf.gfile.FastGFile(image, 'rb').read()
# Creates graph from saved GraphDef.
self.__create_graph()
with tf.Session() as sess:
# 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/contents:0': image_data})
predictions = np.squeeze(predictions)
# Creates node ID --> English string lookup.
node_lookup = NodeLookup()
top_k = predictions.argsort()[-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))
return
def on_partitions_assigned(self, assigned):
print(assigned)
consumer.seek(TopicPartition("Panda_Media", 0), 0)
return
maybe_download_and_extract()
consumer.subscribe(["Panda_Media"], listener=CRL())
# Creates graph from saved GraphDef.
create_graph()
# Creates node ID --> English string lookup.
node_lookup = NodeLookup()
# read messages
for msg in consumer:
# deserialize from json
twete = jsonpickle.decode(msg.value)
# for each media object in tweet
for media in twete.entities['media']:
# base64 jpg string to bytes
image_data = base64.b64decode(media['data'])
# make sure image is jpeg
if is_jpg(image_data) == False:
print("Invalid panda {0}".format(msg.offset))
continue
if request.method == 'POST':
file = request.files['file']
old_file_name = file.filename
if file and allowed_files(old_file_name):
filename = rename_filename(old_file_name)
file_path = os.path.join(UPLOAD_FOLDER, filename)
file.save(file_path)
type_name = 'N/A'
print('file saved to %s' % file_path)
start_time = time.time()
out_html = inference(file_path)
duration = time.time() - start_time
print('duration:[%.0fms]' % (duration*1000))
return result + out_html
return result
##http://127.0.0.1:5001/
if __name__ == "__main__":
print('listening on port %d' % FLAGS.port)
init_graph(model_name=FLAGS.model_name)
label_file = FLAGS.label_file
if not FLAGS.label_file:
label_file, _ = os.path.splitext(FLAGS.model_name)
label_file = label_file + '.label'
node_lookup = NodeLookup(label_file)
app.node_lookup = node_lookup
sess = tf.Session()
app.sess = sess
app.run(host='0.0.0.0', port=FLAGS.port, debug=FLAGS.debug, threaded=True)