opencv\u306f\u3044\u3063\u3066\u306a\u304b\u3063\u305f<\/p>\n
<\/p>\n
http:\/\/qiita.com\/suppy193\/items\/91609e75789e9f458c39<\/p>\n
\u3067Opencv2.7\u3044\u308c\u3066<\/p>\n
http:\/\/arkouji.cocolog-nifty.com\/blog\/2016\/08\/tensorflowraspb.html<\/p>\n
\u3067classfy_image.py\u304c\u306a\u3044\uff08\u305d\u3082\u305d\u3082models\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u304c\u306a\u3044\uff09\u306e\u3067\u3001TensorFlow\u306e\u30bd\u30fc\u30b9\u307f\u3066<\/p>\n
nano classify_image.py<\/p>\n
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.\r\n#\r\n# Licensed under the Apache License, Version 2.0 (the \"License\");\r\n# you may not use this file except in compliance with the License.\r\n# You may obtain a copy of the License at\r\n#\r\n# http:\/\/www.apache.org\/licenses\/LICENSE-2.0\r\n#\r\n# Unless required by applicable law or agreed to in writing, software\r\n# distributed under the License is distributed on an \"AS IS\" BASIS,\r\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\r\n# See the License for the specific language governing permissions and\r\n# limitations under the License.\r\n# ==============================================================================\r\n\r\n\"\"\"Simple image classification with Inception.\r\n\r\nRun image classification with Inception trained on ImageNet 2012 Challenge data\r\nset.\r\n\r\nThis program creates a graph from a saved GraphDef protocol buffer,\r\nand runs inference on an input JPEG image. It outputs human readable\r\nstrings of the top 5 predictions along with their probabilities.\r\n\r\nChange the --image_file argument to any jpg image to compute a\r\nclassification of that image.\r\n\r\nPlease see the tutorial and website for a detailed description of how\r\nto use this script to perform image recognition.\r\n\r\nhttps:\/\/tensorflow.org\/tutorials\/image_recognition\/\r\n\"\"\"\r\n\r\nfrom __future__ import absolute_import\r\nfrom __future__ import division\r\nfrom __future__ import print_function\r\n\r\nimport argparse\r\nimport os.path\r\nimport re\r\nimport sys\r\nimport tarfile\r\n\r\nimport numpy as np\r\nfrom six.moves import urllib\r\nimport tensorflow as tf\r\n\r\nFLAGS = None\r\n\r\n# pylint: disable=line-too-long\r\nDATA_URL = 'http:\/\/download.tensorflow.org\/models\/image\/imagenet\/inception-2015-12-05.tgz'\r\n# pylint: enable=line-too-long\r\n\r\n\r\nclass NodeLookup(object):\r\n \"\"\"Converts integer node ID's to human readable labels.\"\"\"\r\n\r\n def __init__(self,\r\n label_lookup_path=None,\r\n uid_lookup_path=None):\r\n if not label_lookup_path:\r\n label_lookup_path = os.path.join(\r\n FLAGS.model_dir, 'imagenet_2012_challenge_label_map_proto.pbtxt')\r\n if not uid_lookup_path:\r\n uid_lookup_path = os.path.join(\r\n FLAGS.model_dir, 'imagenet_synset_to_human_label_map.txt')\r\n self.node_lookup = self.load(label_lookup_path, uid_lookup_path)\r\n\r\n def load(self, label_lookup_path, uid_lookup_path):\r\n \"\"\"Loads a human readable English name for each softmax node.\r\n\r\n Args:\r\n label_lookup_path: string UID to integer node ID.\r\n uid_lookup_path: string UID to human-readable string.\r\n\r\n Returns:\r\n dict from integer node ID to human-readable string.\r\n \"\"\"\r\n if not tf.gfile.Exists(uid_lookup_path):\r\n tf.logging.fatal('File does not exist %s', uid_lookup_path)\r\n if not tf.gfile.Exists(label_lookup_path):\r\n tf.logging.fatal('File does not exist %s', label_lookup_path)\r\n\r\n # Loads mapping from string UID to human-readable string\r\n proto_as_ascii_lines = tf.gfile.GFile(uid_lookup_path).readlines()\r\n uid_to_human = {}\r\n p = re.compile(r'[n\\d]*[ \\S,]*')\r\n for line in proto_as_ascii_lines:\r\n parsed_items = p.findall(line)\r\n uid = parsed_items[0]\r\n human_string = parsed_items[2]\r\n uid_to_human[uid] = human_string\r\n\r\n # Loads mapping from string UID to integer node ID.\r\n node_id_to_uid = {}\r\n proto_as_ascii = tf.gfile.GFile(label_lookup_path).readlines()\r\n for line in proto_as_ascii:\r\n if line.startswith(' target_class:'):\r\n target_class = int(line.split(': ')[1])\r\n if line.startswith(' target_class_string:'):\r\n target_class_string = line.split(': ')[1]\r\n node_id_to_uid[target_class] = target_class_string[1:-2]\r\n\r\n # Loads the final mapping of integer node ID to human-readable string\r\n node_id_to_name = {}\r\n for key, val in node_id_to_uid.items():\r\n if val not in uid_to_human:\r\n tf.logging.fatal('Failed to locate: %s', val)\r\n name = uid_to_human[val]\r\n node_id_to_name[key] = name\r\n\r\n return node_id_to_name\r\n\r\n def id_to_string(self, node_id):\r\n if node_id not in self.node_lookup:\r\n return ''\r\n return self.node_lookup[node_id]\r\n\r\n\r\ndef create_graph():\r\n \"\"\"Creates a graph from saved GraphDef file and returns a saver.\"\"\"\r\n # Creates graph from saved graph_def.pb.\r\n with tf.gfile.FastGFile(os.path.join(\r\n FLAGS.model_dir, 'classify_image_graph_def.pb'), 'rb') as f:\r\n graph_def = tf.GraphDef()\r\n graph_def.ParseFromString(f.read())\r\n _ = tf.import_graph_def(graph_def, name='')\r\n\r\n\r\ndef run_inference_on_image(image):\r\n \"\"\"Runs inference on an image.\r\n\r\n Args:\r\n image: Image file name.\r\n\r\n Returns:\r\n Nothing\r\n \"\"\"\r\n if not tf.gfile.Exists(image):\r\n tf.logging.fatal('File does not exist %s', image)\r\n image_data = tf.gfile.FastGFile(image, 'rb').read()\r\n\r\n # Creates graph from saved GraphDef.\r\n create_graph()\r\n\r\n with tf.Session() as sess:\r\n # Some useful tensors:\r\n # 'softmax:0': A tensor containing the normalized prediction across\r\n # 1000 labels.\r\n # 'pool_3:0': A tensor containing the next-to-last layer containing 2048\r\n # float description of the image.\r\n # 'DecodeJpeg\/contents:0': A tensor containing a string providing JPEG\r\n # encoding of the image.\r\n # Runs the softmax tensor by feeding the image_data as input to the graph.\r\n softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')\r\n predictions = sess.run(softmax_tensor,\r\n {'DecodeJpeg\/contents:0': image_data})\r\n predictions = np.squeeze(predictions)\r\n\r\n # Creates node ID --> English string lookup.\r\n node_lookup = NodeLookup()\r\n\r\n top_k = predictions.argsort()[-FLAGS.num_top_predictions:][::-1]\r\n for node_id in top_k:\r\n human_string = node_lookup.id_to_string(node_id)\r\n score = predictions[node_id]\r\n print('%s (score = %.5f)' % (human_string, score))\r\n\r\n\r\ndef maybe_download_and_extract():\r\n \"\"\"Download and extract model tar file.\"\"\"\r\n dest_directory = FLAGS.model_dir\r\n if not os.path.exists(dest_directory):\r\n os.makedirs(dest_directory)\r\n filename = DATA_URL.split('\/')[-1]\r\n filepath = os.path.join(dest_directory, filename)\r\n if not os.path.exists(filepath):\r\n def _progress(count, block_size, total_size):\r\n sys.stdout.write('\\r>> Downloading %s %.1f%%' % (\r\n filename, float(count * block_size) \/ float(total_size) * 100.0))\r\n sys.stdout.flush()\r\n filepath, _ = urllib.request.urlretrieve(DATA_URL, filepath, _progress)\r\n print()\r\n statinfo = os.stat(filepath)\r\n print('Successfully downloaded', filename, statinfo.st_size, 'bytes.')\r\n tarfile.open(filepath, 'r:gz').extractall(dest_directory)\r\n\r\n\r\ndef main(_):\r\n maybe_download_and_extract()\r\n image = (FLAGS.image_file if FLAGS.image_file else\r\n os.path.join(FLAGS.model_dir, 'cropped_panda.jpg'))\r\n run_inference_on_image(image)\r\n\r\n\r\nif __name__ == '__main__':\r\n parser = argparse.ArgumentParser()\r\n # classify_image_graph_def.pb:\r\n # Binary representation of the GraphDef protocol buffer.\r\n # imagenet_synset_to_human_label_map.txt:\r\n # Map from synset ID to a human readable string.\r\n # imagenet_2012_challenge_label_map_proto.pbtxt:\r\n # Text representation of a protocol buffer mapping a label to synset ID.\r\n parser.add_argument(\r\n '--model_dir',\r\n type=str,\r\n default='\/tmp\/imagenet',\r\n help=\"\"\"\\\r\n Path to classify_image_graph_def.pb,\r\n imagenet_synset_to_human_label_map.txt, and\r\n imagenet_2012_challenge_label_map_proto.pbtxt.\\\r\n \"\"\"\r\n )\r\n parser.add_argument(\r\n '--image_file',\r\n type=str,\r\n default='',\r\n help='Absolute path to image file.'\r\n )\r\n parser.add_argument(\r\n '--num_top_predictions',\r\n type=int,\r\n default=5,\r\n help='Display this many predictions.'\r\n )\r\n FLAGS, unparsed = parser.parse_known_args()\r\n tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)\r\n<\/pre>\n\u3067\u3001<\/p>\n
python classify_image.py<\/p>\n
\u3084\u3063\u3066<\/p>\n
giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca (score = 0.89107)
\nindri, indris, Indri indri, Indri brevicaudatus (score = 0.00779)
\nlesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens (score = 0.00296)
\ncustard apple (score = 0.00147)
\nearthstar (score = 0.00117)<\/p>\n
\u30d1\u30f3\u30c089%\u3000\u30a4\u30f3\u30c9\u30ea\uff08\u30b5\u30eb\uff090.7%\u3001\u30ec\u30c3\u30b5\u30fc\u30d1\u30f3\u30c00.2%\u3001custard apple\uff08\u30ea\u30f3\u30b4\u306e\u4ef2\u9593\uff1f\uff090.14\uff05\u3001\u30a2\u30fc\u30b9\u30b9\u30bf\u30fc\u3063\u3066\u8b0e\u306e\u82b10.11%\u3000\u3058\u3083\u306a\u3044\u3063\u3066\u51fa\u308b<\/p>\n
<\/p>\n
\u81ea\u5206\u3067\u4f5c\u308b\u53c2\u8003<\/p>\n
http:\/\/arkouji.cocolog-nifty.com\/blog\/2016\/08\/tensorflow-76e9.html<\/p>\n
\u3053\u306e\u30b5\u30a4\u30c8\u3068\u305d\u3053\u306b\u306e\u3063\u3066\u308b\u53c2\u8003\u30b5\u30a4\u30c8<\/p>\n
http:\/\/qiita.com\/khayate\/items\/bb7c61f447b4c579ddd1<\/p>\n
<\/p>\n
\u308f\u304b\u308a\u3084\u3059\u3044\u89e3\u8aac<\/p>\n