{"id":2529,"date":"2022-09-29T11:00:00","date_gmt":"2022-09-29T02:00:00","guid":{"rendered":"https:\/\/media.value-domain.com\/?p=2529"},"modified":"2022-11-21T11:30:07","modified_gmt":"2022-11-21T02:30:07","slug":"gcp-tensorflow","status":"publish","type":"post","link":"https:\/\/media.value-domain.com\/gcp-tensorflow\/","title":{"rendered":"GCP AutoML Vison\u306e\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3068TensorFlow\u3067AI\u753b\u50cf\u63a8\u8ad6\u3059\u308b"},"content":{"rendered":"

\u306f\u3058\u3081\u306b<\/h2>\n\n\n

\u524d\u56de\u3001Google AutoML Vision<\/a>\u3092\u4f7f\u3063\u3066\u300cAI\u753b\u50cf\u89e3\u6790\u3067\u30a4\u30e1\u30fc\u30b8\u30ab\u30e9\u30fc\u3092\u62bd\u51fa\u3059\u308b<\/a>\u300d\u3092\u7d39\u4ecb\u3057\u307e\u3057\u305f\u3002(\u624b\u9806\u8a73\u7d30\u306f\u3053\u3061\u3089<\/a>)
\"\"
<\/p>\n\n\n\n

Google\u306eAutoML Vision\u306f\u3042\u3089\u304b\u3058\u3081\u7528\u610f\u3055\u308c\u3066\u3044\u308b\u5b66\u7fd2\u6e08\u307f\u306e\u753b\u50cf\u8a8d\u8b58\u30e2\u30c7\u30eb\u3092\u5229\u7528\u3057\u3001\u305d\u306e\u30e2\u30c7\u30eb\u3092\u571f\u53f0\u3068\u3057\u3066\u4e00\u90e8\u3092\u5909\u66f4\u3059\u308b\u3053\u3068\u3067\u3001\u65b0\u3057\u3044\u554f\u984c\u306b\u5bfe\u5fdc\u3067\u304d\u308b\u6a5f\u68b0\u5b66\u7fd2\u30b5\u30fc\u30d3\u30b9\u3067\u3059\u3002
\u3053\u306e\u5b66\u7fd2\u65b9\u6cd5\u306f\u8ee2\u79fb\u5b66\u7fd2\u3068\u547c\u3070\u308c\u307e\u3059\u3002\u30e6\u30fc\u30b6\u306f\u30e9\u30d9\u30eb\u4ed8\u304d\u753b\u50cf\u30c7\u30fc\u30bf\u3092AutoML Vision\u306b\u6295\u5165\u3059\u308b\u3060\u3051\u3067\u3059\u3002AutoML Vision\u306f\u305d\u306e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u5143\u306b\u300c\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u300d\u3092\u5b9f\u884c\u3057\u3001\u672a\u77e5\u306e\u30d1\u30bf\u30fc\u30f3\u3092\u8a8d\u8b58\u3059\u308b\u3088\u3046\u5b66\u7fd2\u3059\u308b\u3002
\u30e6\u30fc\u30b6\u306f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3055\u3048\u7528\u610f\u3067\u304d\u308c\u3070\u30bc\u30ed\u304b\u3089\u753b\u50cf\u8a8d\u8b58\u30e2\u30c7\u30eb\u3092\u958b\u767a\u3059\u308b\u5fc5\u8981\u304c\u306a\u304f\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n

\u4eca\u56de\u306f\u3053\u306eGoogle Cloud AutoML Vision\u3067\u751f\u6210\u3057\u305f\u65b0\u305f\u306a\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3092cloud\u4e0a\u306b\u30c7\u30d7\u30ed\u30a4\u305b\u305a\u306bPC\u7aef\u672b\u306b\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u3057\u3066\u3001\u30d7\u30ed\u30b0\u30e9\u30df\u30f3\u30b0\u8a00\u8a9e\u306ePython<\/strong>\u3068\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u306eTensorFlow<\/strong>\u3092\u7528\u3044\u3066PC\u7aef\u672b\u5074\u3067\u63a8\u8ad6\u3057\u307e\u3059\u3002cloud\u4e0a\u3067\u63a8\u8ad6\u305b\u305a\u81ea\u8eab\u306ePC\u3084IoT\u6a5f\u5668\u5358\u72ec\u3067\u63a8\u8ad6\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002
\u3000\u203bTensorFlow \u306fGoogle\u304c\u958b\u767a\u3057\u305f\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u306e\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u3067\u3059\u3002
TensorFlow\u306e\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u306e\u30d5\u30a1\u30a4\u30eb\u3067\u3042\u308bsaved_model.pb<\/strong>\u3092\u7528\u610f\u6e08\u307f\u3067\u3001\u5177\u4f53\u7684\u306a\u624b\u6cd5\u3092\u53c2\u7167\u3055\u308c\u305f\u3044\u65b9\u306f\u30015.TensorFlow\u30e2\u30c7\u30eb\u306esaved_model.pb<\/strong>\u304b\u3089\u8aad\u307f\u9032\u3081\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n

<\/p>\n\n\n\n

Auto ML\u306eCloud\u30db\u30b9\u30c8\u3068edge\u30e2\u30fc\u30c9<\/h2>\n\n\n\n


Google Cloud AutoML Vision\u306b\u306fcloud\u30db\u30b9\u30c8\u3068edge\u30e2\u30fc\u30c9\u306e\uff12\u7a2e\u985e\u304c\u3042\u308a\u307e\u3059\u3002cloud\u30db\u30b9\u30c8\u306f\u5b66\u7fd2\uff08TRAIN\uff09\u3001\u8a55\u4fa1\uff08EVALUTE\uff09\u3001\u30c6\u30b9\u30c8\uff08TEST&USE\uff09\u3001\u63a8\u8ad6\uff08INTERFACE\uff09\u5168\u3066\u3092\u30af\u30e9\u30a6\u30c9\u4e0a\u3067\u884c\u3044\u307e\u3059\u3002\u307e\u305fAPI\u9023\u643a\u306b\u3088\u3063\u3066\u30aa\u30f3\u30e9\u30a4\u30f3\u3067\u63a8\u8ad6\u3082\u53ef\u80fd\u3067\u3059\u3002
\u305d\u308c\u306b\u5bfe\u3057\u3066edge\u30e2\u30fc\u30c9\u306f\u5b66\u7fd2\uff08TRAIN\uff09\u3001\u8a55\u4fa1\uff08EVALUTE\uff09\u3092\u884c\u3044\u5b66\u7fd2\u6e08\u307f\u30c7\u30fc\u30bf\u306e\u751f\u6210\u306f\u30d5\u30a7\u30fc\u30ba\u307e\u3067\u3068\u306a\u308a\u307e\u3059\u3002\u63a8\u8ad6\u306f\u3053\u306e\u5b66\u7fd2\u6e08\u307f\u30c7\u30fc\u30bf\u3092\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u3057\u3066PC\u3084IoT\u6a5f\u5668\u5074\u3067TensorFlow\u3092\u7528\u3044\u3066\u5b9f\u65bd\u3067\u304d\u307e\u3059\u3002
\u30af\u30e9\u30a6\u30c9AI\uff08Cloud AI\uff09\u306b\u5bfe\u3057\u3066\u7aef\u672b\u5074\u3067\u63a8\u8ad6\u3059\u308b\u65b9\u6cd5\u3068\u3057\u3066\u30a8\u30c3\u30b8AI(Edge AI)\u3068\u8a00\u308f\u308c\u307e\u3059\u3002IoT\u7b49\u306e\u7aef\u672b\u5074\u3067\u63a8\u8ad6\u3059\u308b\u3053\u3068\u3067\u30ea\u30a2\u30eb\u30bf\u30a4\u30e0\u306a\u5224\u65ad\u304c\u3067\u304d\u307e\u3059\u3002\u305f\u3060\u3057\u3001\u7aef\u672b\u5074\u3067\u884c\u3046\u306e\u3067\u305d\u306e\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u306f\u8efd\u91cf\u306a\u3082\u306e\u3067\u3001\u9ad8\u5ea6\u306a\u8a08\u7b97\u30fb\u5224\u65ad\u306f\u96e3\u3057\u3044\u30c7\u30e1\u30ea\u30c3\u30c8\u304c\u3042\u308a\u307e\u3059\u3002
\u524d\u56de\u306f\u30e2\u30c7\u30eb\u306e\u5b9a\u7fa9\u3068\u63a8\u8ad6\u306fcloud\u30db\u30b9\u30c8\u3067\u884c\u3063\u3066\u3044\u307e\u3057\u305f\u3002\u753b\u50cf\u8a8d\u8b58\u30e2\u30c7\u30eb\u306e\u5834\u5408\u306f\u5b66\u7fd2\u6e08\u307f\u30c7\u30fc\u30bf\u3092\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u3067\u304d\u307e\u305b\u3093\u3002edge\u30e2\u30fc\u30c9\u3067\u5b66\u7fd2\u6e08\u307f\u30c7\u30fc\u30bf\u3092\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n

cloud\u30db\u30b9\u30c8\u3067\u5b9f\u65bd[Fig1]\u3057\u305f\u3053\u3068\u3092\u518d\u5ea6\u3001edge\u30e2\u30c7\u30eb\u3067\u3084\u308a\u76f4\u3057\u307e\u3057\u305f\u3002edge\u30e2\u30c7\u30eb\u3067\u306f\u6700\u9069\u5316\u306b\u9ad8\u901f\u3001\u4e2d\u9593\u3001\u9ad8\u7cbe\u5ea6\u306e\uff13\u7a2e\u985e\u3042\u308a\u9ad8\u7cbe\u5ea6\uff08Higher accuracy\uff09\u3067\u30e2\u30c7\u30eb\u4f5c\u6210\u3057\u307e\u3057\u305f\u3002[Fig2]<\/p>\n\n\n\n

\"\"<\/p>\n\n\n\n

edge\u30e2\u30fc\u30c9\u3067\u306e\u7d50\u679c\u306f\u524d\u56de\u306ecloud\u30db\u30b9\u30c8\u3068\u307b\u307c\u540c\u69d8\u306e\u7d50\u679c\u3067\u3059\u3002\u307b\u3093\u306e\u50c5\u304b\u306b\u9069\u5408\u7387\u304c\u4e0a\u304c\u3063\u3066\u3044\u307e\u3059\u3002
\u203bAutoML Vision \u521d\u5fc3\u8005\u5411\u3051\u30ac\u30a4\u30c9\u306b\u8a55\u4fa1\u7d50\u679c\u306e\u898b\u65b9\u306e\u89e3\u8aac\u304c\u3042\u308a\u307e\u3059\u306e\u3067\u53c2\u8003\u306b\u3057\u3066\u304f\u3060\u3055\u3044\u3002
https:\/\/cloud.google.com\/vision\/automl\/docs\/beginners-guide?hl=ja<\/a><\/p>\n\n\n\n

cloud\u30db\u30b9\u30c8\u30e2\u30fc\u30c9\u3067\u5b9f\u65bd\u3057\u305f\u3068\u304d\u306f\u3001139.964\u30ce\u30fc\u30c9\u6642\u9593<\/strong>\u3092\u8cbb\u3084\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u7528\u30af\u30ec\u30b8\u30c3\u30c8\u3068\u7121\u6599\u30af\u30ec\u30b8\u30c3\u30c8\u3092\u4f7f\u3044\u679c\u305f\u3057\u3001\u3055\u3089\u306b6,000\u5186\u307b\u3069\u5fb4\u53ce\u3055\u308c\u307e\u3057\u305f\u3002(1$=114.9\u5186 2022\/3\u6642\u70b9) [Fig3]
edge\u30e2\u30fc\u30c9\u3067\u5b9f\u65bd\u3057\u305f\u5834\u5408\u306f2.41\u30ce\u30fc\u30c9\u6642\u9593<\/strong>\u3067\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6642\u9593\u3082\uff14\u6642\u9593\u7a0b\u5ea6\u3067\u6e08\u307f\u307e\u3057\u305f\u3002\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u7528\u306e\u30af\u30ec\u30b8\u30c3\u30c8\u7121\u6599\u5206\u3067$42\u6d88\u8cbb\u3057\u305f\u3060\u3051\u3067\u3001\u307e\u3060\u4f59\u88d5\u3067\u7121\u6599\u30af\u30ec\u30b8\u30c3\u30c8\u304c\u6b8b\u3063\u3066\u3044\u307e\u3059\u3002(1$=136.9\u5186 2022\/8\u6642\u70b9) [Fig4]
\"\" <\/p>\n\n\n\n

\u203b\u65b0\u898f\u306bGoogle Cloud Platform(GCP)\u3092\u4f7f\u3046\u5834\u5408\u306f\u7121\u6599\u30c8\u30e9\u30a4\u30a2\u30eb\u3068\u3057\u3066$300\u3001\u4f01\u696d\u30a2\u30ab\u30a6\u30f3\u30c8\u3068\u3057\u3066\u8a8d\u5b9a\u3055\u308c\u305f\u5834\u5408\u306f\u8ffd\u52a0\u3067100$\u3001\u66f4\u306b\u521d\u3081\u3066AutoML\u3092\u4f7f\u3046\u5834\u5408$50\uff08\u30aa\u30f3\u30e9\u30a4\u30f3\u5229\u7528\uff09\u3001$75\uff08\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\uff09\u304c\u4ed8\u4e0e\u3055\u308c\u308b\u306e\u3067\u5341\u5206\u7121\u6599\u3067\u4f5c\u6210\u53ef\u80fd\u3067\u3059\u3002
\u203b\u3053\u3053\u534a\u5e74\u3067\u76f8\u5f53\u306a\u5186\u5b89\u306b\u306a\u3063\u3066\u304a\u308a\u3001\u30c9\u30eb\u3067\u7121\u6599\u30af\u30ec\u30b8\u30c3\u30c8\u3092\u53d6\u5f97\u3059\u308b\u3068\u3059\u3054\u304f\u5f97\u3057\u3066\u3044\u308b\u3088\u3046\u306b\u898b\u3048\u307e\u3059\u304c\u3001\u30ce\u30fc\u30c9\u6642\u9593<\/strong>\u306f\u30c9\u30eb\u8ab2\u91d1\u306a\u306e\u3067\u5909\u308f\u308a\u306f\u3042\u308a\u307e\u305b\u3093\u3002\u305f\u3060\u3057\u4f01\u696d\u3067\u5186\u3067\u8ab2\u91d1\u3059\u308b\u5834\u5408\u306f\u6050\u308d\u3057\u304f\u5024\u4e0a\u3052\u3068\u306a\u308a\u307e\u3059\u3002\u5927\u91cf\u306e\u30ce\u30fc\u30c9\u6642\u9593\u3092\u5fc5\u8981\u3068\u3059\u308b\u3088\u3046\u306a\u5834\u5408\u306f\u5927\u5e45\u306a\u5024\u4e0a\u3052\u3068\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n

\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u306e\u30a8\u30af\u30b9\u30dd\u30fc\u30c8<\/h2>\n\n\n

Fig2.\u306eAutoML Vision edge\u30e2\u30fc\u30c9\u3067\u5b66\u7fd2\u3057\u305f\u30e2\u30c7\u30eb\u306f\u3001\u300c\u30c6\u30b9\u30c8\u3068\u4f7f\u7528\u300d\u304b\u3089\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u304c\u9078\u629e\u3067\u304d\u307e\u3059\u3002
TensorFlow\u3067\u63a8\u8ad6\u3092\u5b9f\u65bd\u3057\u305f\u3044\u5834\u5408\u306f\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30ebsaved_model.pb\u3092\u5229\u7528\u3057\u307e\u3059\u3002\u300ccontainer\u300d\u3092\u9078\u629e\u3059\u308c\u3070 TensorFlow\u30e2\u30c7\u30eb\u7528\u306esaved_model.pb\u3092\u53d6\u5f97\u3067\u304d\u307e\u3059\u3002[Fig5]
\u300ccontainer\u300d\u306fGoogle\u304c\u7528\u610f\u3057\u3066\u304f\u308c\u305f\u5c02\u7528\u306eDocker(\u30b3\u30f3\u30c6\u30ca\u578b\u306e\u4eee\u60f3\u74b0\u5883)\u7528\u3067\u3059\u304c\u3001Docker\u306f\u4f7f\u7528\u305b\u305a\u72ec\u81ea\u306e\u5b9f\u884c\u74b0\u5883\u3067TensorFlow\u3092\u4f7f\u3063\u3066\u691c\u8a3c\u3067\u304d\u307e\u3059\u3002
\"\"

\u30a4\u30e1\u30fc\u30b8\u30ab\u30e9\u30fc\u5b66\u7fd2\u6e08\u307f\u306e\u30e2\u30c7\u30ebsaved_model.pb\u306f23M\u30d0\u30a4\u30c8\u7a0b\u5ea6\u3067\u3059\u3002\u9ad8\u7cbe\u5ea6\u3067\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u307e\u3057\u305f\u304c\u3001cloud\u30db\u30b9\u30c8\u306e139.964\u30ce\u30fc\u30c9\u6642\u9593\u306b\u5bfe\u3057\u3066\u50c5\u304b2.41\u30ce\u30fc\u30c9\u6642\u9593\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3068\u306a\u3063\u3066\u3044\u308b\u305f\u3081\u7aef\u672b\u5411\u3051\u306b\u30d5\u30a1\u30a4\u30eb\u30b5\u30a4\u30ba\u304c\u5c0f\u3055\u304f\u306a\u3063\u3066\u3044\u307e\u3059\u3002\u6b21\u306e\u30bb\u30af\u30b7\u30e7\u30f3\u3067\u3053\u306esaved_model.pb\u3092\u5229\u7528\u3057\u3066\u5b9f\u969b\u306b\u63a8\u8ad6\u3092\u3057\u3066\u3044\u304d\u307e\u3059\u3002\u305d\u306e\u70ba\u306bpython\u3068tensorflow\u306e\u30e9\u30a4\u30d6\u30ea\u30fc\u304c\u5fc5\u8981\u3068\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n

python\u3068TensorFlow\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb<\/h2>\n\n\n

https:\/\/www.python.org\/<\/a>\u304b\u3089python-3.10.6-amd64.exe(windows\u7248)\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3057\u3066\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\u3002\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u5f8c\u3001cmd\u753b\u9762\u304b\u3089pip\u30b3\u30de\u30f3\u30c9\u3067\u5fc5\u8981\u306a\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n

\n
python.exe -m pip install --upgrade pip\npip3 install numpy\npip3 install opencv-python\npip3 install tensorflow\npip3 install argparse<\/pre>\n<\/div>\n\n\n\n
\u203b\u63a8\u8ad6\u3060\u3051\u3067\u3059\u306e\u3067NIVIDA\u7cfb\u306eGPU\u304c\u306a\u304f\u3066\u3082\u8a08\u7b97\u7d50\u679c\u306f\u3059\u3050\u306b\u3067\u307e\u3059\u3002<\/p>\n\n\n
TensorFlow\u30e2\u30c7\u30eb\u306esaved_model.pb<\/h2>\n\n\n
AutoML Vision edge\u30e2\u30fc\u30c9\u3067\u4f5c\u6210\u3057\u305f\u753b\u50cf\u8a8d\u8b58\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u306esaved_model.pb\u306f\u3001\u3069\u306e\u3088\u3046\u306a\u69cb\u9020\u306a\u306e\u304b\u3002\u307e\u305f\u63a8\u8ad6\u3057\u305f\u30c7\u30fc\u30bf\u306f\u3069\u3046\u8a2d\u5b9a\u3059\u308c\u3070\u826f\u3044\u306e\u304b\u8abf\u3079\u3066\u307f\u307e\u3059\u3002
tensorflow\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u3068saved_model_cli<\/strong>\u30b3\u30de\u30f3\u30c9\u304c\u5229\u7528\u3067\u304d\u3001TensorFlow\u306eSaved Model\u3092\u78ba\u8a8d\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n
\n
c:\\apps\\ai>saved_model_cli show --dir . --all\n\nMetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:\nsignature_def['serving_default']:\nThe given SavedModel SignatureDef contains the following input(s):\ninputs['image_bytes'] tensor_info:\ndtype: DT_STRING\nshape: (-1)\nname: Placeholder:0\ninputs['key'] tensor_info:\ndtype: DT_STRING\nshape: (-1)\nname: Placeholder_1:0\nThe given SavedModel SignatureDef contains the following output(s):\noutputs['key'] tensor_info:\ndtype: DT_STRING\nshape: (-1)\nname: Identity:0\noutputs['labels'] tensor_info:\ndtype: DT_STRING\nshape: (-1, 14)\nname: Tile:0\noutputs['scores'] tensor_info:\ndtype: DT_FLOAT\nshape: (-1, 14)\nname: scores:0\nMethod name is: tensorflow\/serving\/predict\n<\/pre>\n<\/div>\n\n\n\n
\u30fbinput\u30c7\u30fc\u30bf\u5185\u5bb9 
 \u3000\u2460image_bytes(DT_STRING\u578b) \u63a8\u8ad6\u3057\u305f\u3044\u753b\u50cf\u30c7\u30fc\u30bf 
 \u3000\u2461key(DT_STRING\u578b) \u63a8\u8ad6\u3059\u308b\u3068\u304d\u306e\u30e6\u30cb\u30fc\u30af\u30ad\u30fc 
\u30fboutput\u30c7\u30fc\u30bf\u5185\u5bb9
 \u3000\u2462key(DT_STRING\u578b) \u63a8\u8ad6\u7d50\u679c\u5fdc\u7b54\u30ad\u30fc\u2461\u3092\u8fd4\u3059 
 \u3000\u2463labels(DT_STRING\u578b) \u8272\u540d \u3000 (\u30a4\u30e1\u30fc\u30b8\u30ab\u30e9\u30fc<\/a>\u3092\u5b66\u7fd2\u3055\u305b\u305f\u306e\u3067\u3001\u8272\u540d\u306e\u30e9\u30d9\u30eb\u4e00\u89a7\u3092\u8fd4\u3059)
 \u3000\u2464scores(DT_FLOAT\u578b) \u9069\u5fdc\u7387\u3000(\u5404\u30e9\u30d9\u30eb\u306b\u5bfe\u3059\u308b\u9069\u5fdc\u7387\u3092\u8fd4\u3059)<\/p>\n\n\n\n
\u2461\u306ekey(DT_STRING\u578b) \u306f\u30e6\u30cb\u30fc\u30af\u6587\u5b57\u5217\u3067\u306a\u3093\u3067\u3082\u304b\u307e\u3044\u307e\u305b\u3093\u3002\u3053\u308c\u306f\u8981\u6c42\u3055\u308c\u305f\u30bb\u30c3\u30b7\u30e7\u30f3\u306b\u5bfe\u3059\u308b\u5fdc\u7b54\u3067\u3042\u308b\u304b\u3092\u5224\u65ad\u3059\u308b\u305f\u3081\u3067\u3001\u975e\u540c\u671f\u306b\u8981\u6c42\u3068\u7d50\u679c\u3092\u5bfe\u5fdc\u3055\u305b\u308b\u3068\u304d\u306b\u3088\u304f\u4f7f\u3046\u624b\u6bb5\u3067\u4eca\u56de\u306f\u30ef\u30f3\u30b7\u30e7\u30c3\u30c8\u306a\u306e\u3067\u6c17\u306b\u3057\u307e\u305b\u3093\u3002<\/p>\n\n\n\n
\u554f\u984c\u306f\u2460\u306eimage_bytes\u304cDT_STRING\u578b\u306b\u306a\u3063\u3066\u3044\u308b\u3053\u3068\u3067\u3059\u3002\u753b\u50cf\u30c7\u30fc\u30bf\u3092\u6e21\u3059\u3053\u3068\u306b\u306a\u308a\u307e\u3059\u304c\u3001\u6587\u5b57\u5217\u578b\u3068\u306f\u3069\u3046\u3044\u3046\u6271\u3044\u3067\u3057\u3087\u3046\u304b\uff1f \u30d0\u30a4\u30ca\u30ea\u30fc\u3092\uff28\uff25\uff38\u6587\u5b57\u5217\u3067\u8868\u3059\u306e\u3067\u3057\u3087\u3046\u304b\uff1f 
AutoML Vision edge\u30e2\u30fc\u30c9\u3067\u300ccontainer\u300d\u306fDocker\u3067\u306e\u5229\u7528\u3092\u524d\u63d0\u3068\u3057\u3066\u3044\u307e\u3059\u3002python\u3067\u306e\u30b3\u30fc\u30c9\u304c\u4ee5\u4e0b\u306b\u8a18\u8f09\u304c\u3042\u308a\u307e\u3059\u3002 
https:\/\/cloud.google.com\/vision\/automl\/docs\/containers-gcs-tutorial?hl=ja#container-predict-example-python<\/a><\/p>\n\n\n\n
instances = {\n        'instances': [\n                {'image_bytes': {'b64': str(encoded_image)},\n                 'key': image_key}\n        ]\n}<\/code><\/pre>\n\n\n\n
image_bytes\u306fBASE64\u3067\u6587\u5b57\u5217\u578b\u306b\u3057\u3066\u3044\u308b\u3053\u3068\u304c\u308f\u304b\u308a\u307e\u3059\u304c\u3001\u3053\u306e\u3068\u304a\u308a\u306e\u6587\u5b57\u5217\u3067\u8a66\u3057\u3066\u307f\u307e\u3057\u305f\u304c\u3060\u3081\u3067\u3057\u305f\u3002\u3053\u308c\u306fDocker\u7528\u306b\u7d44\u307f\u8fbc\u307e\u308c\u305f\u3082\u306e\u3067\u3001\u76f4\u63a5\u5229\u7528\u306e\u5834\u5408\u306f\u9055\u3044byte\u5217\u306e\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002\u4f8b\u3048\u30702x2\u30d4\u30af\u30bb\u30eb\u767d\u8272\u753b\u50cf\u306epng\u5f62\u5f0f\u306e\u30d0\u30a4\u30ca\u30ea\u30c7\u30fc\u30bf\u306f\uff17\uff16\u30d0\u30a4\u30c8\u3067\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002 [Fig6]
 \"\"<\/p>\n\n\n\n
python\u306ebyte\u5217\u3067\u306f\u4ee5\u4e0b\u306e\u3088\u3046\u306bb\u304c\u4ed8\u304fstring\u578b\u306b\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n
b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIHDR\\x00\\x00\\x00\\x02\\x00\\x00\\x00\\x02\\x08\\x02\\x00\\x00\\x00\\xfd\\xd4\\x9as\\x00\\x00\\x00\\x16IDAT\\x08\\x1dc\\xfc\\xff\\xff?\\x03\\x03\\x03\\xe3\\xff\\xff\\xff\\x19\\x18\\x18\\x002\\xf0\\x05\\xfd\\xc0Eu\\xda\\x00\\x00\\x00\\x00IEND\\xaeB`\\x82'<\/code><\/pre>\n\n\n\n
\u30a4\u30e1\u30fc\u30b8\u30c7\u30fc\u30bf\u3092byte\u5217\u3067\u6e21\u3059\u3060\u3051\u3067\u826f\u3044\u3002\u6b21\u306e\u30bb\u30af\u30b7\u30e7\u30f3\u3067\u5b9f\u969b\u306b\u30d7\u30ed\u30b0\u30e9\u30df\u30f3\u30b0\u3057\u3066\u63a8\u8ad6\u3092\u884c\u3044\u307e\u3059\u3002<\/p>\n\n\n
saved_model.pb\u3092\u30ed\u30fc\u30c9\u3057\u3066\u63a8\u8ad6<\/h2>\n\n\n
python\u3067\u30b3\u30fc\u30c9\u306f\u305f\u3063\u305f\uff13\uff10\u6570\u884c\u3067\u3059\u3002\u4f7f\u3044\u65b9\u306f\u3001python tfcolor2.py \u753b\u50cf\u30d5\u30a1\u30a4\u30eb\u540d1 [\u753b\u50cf\u30d5\u30a1\u30a4\u30eb\u540d2 \u2026]\u3067\u30a4\u30e1\u30fc\u30b8\u30ab\u30e9\u30fc\u30c8\u30c3\u30d7\uff13\u304c\u7d50\u679c\u3068\u3057\u3066\u8fd4\u3055\u308c\u307e\u3059\u3002 
tycolor2.py<\/strong> \u30bd\u30fc\u30b9\u30b3\u30fc\u30c9<\/a><\/p>\n\n\n
import cv2\nimport tensorflow as tf\nimport numpy as np\nimport argparse\n\nparser = argparse.ArgumentParser(description='')\nparser.add_argument('image_path', nargs='+', help='path to image file list')\nargs = parser.parse_args()\nimage_path = args.image_path\n\nloaded = tf.saved_model.load(export_dir='.\/')\nprint(list(loaded.signatures.keys()))\n\ninfer = loaded.signatures[\"serving_default\"]\nprint(infer.structured_outputs)\n\nimgbytes = list()\nfor path in image_path:\n        img = cv2.imread(path)\n        rst,bts = cv2.imencode('.png', img)\n        imgbytes.append(bts.tobytes())\n\nout = infer(key=tf.constant('uniq_dummy_id_123'), image_bytes=tf.constant(imgbytes))\nprint(out)\n\nscores = out[\"scores\"]\nlabels = out[\"labels\"]\n\nnum = scores.numpy().shape[0]\nfor j in range(num):\n        top = np.argsort(-scores.numpy()[j])\n        print(image_path[j])\n        for i in range(3):\n                print('%d: %-7.6f : %s' % (i + 1,scores.numpy()[j][top[i]], labels.numpy()[j][top[i]].decode('utf-8')))\n<\/code><\/pre>\n\n\n
\u30fb6\uff5e9\u884c\uff1a\u753b\u50cf\u30d5\u30a1\u30a4\u30eb\u3092\u8907\u6570\u6307\u5b9a\u3067\u304d\u308b\u3088\u3046\u306b\u3001argparse\u3067\u5f15\u6570\u3092\u53d6\u5f97\u3057\u3066\u3044\u307e\u3059\u3002 
\u30fb11\u884c\uff1atf.saved_model.load\u3067\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30ebsaved_model.pb\u3092\u30ed\u30fc\u30c9\u3057\u307e\u3059\u3002\u30d5\u30a1\u30a4\u30eb\u540d\u306f\u56fa\u5b9a\u3067\u3001saved_model.pb\u304c\u5b58\u5728\u3059\u308b \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u30fc\u3092\u30d1\u30e9\u30e1\u30fc\u30bf\u306b\u6307\u5b9a\u3057\u307e\u3059\u3002 
\u30fb14\u884c\uff1aloaded.signatures[\"serving_default\"]\u3067dict\u578b(\u8f9e\u66f8\u306e\u578b)\u3068\u3057\u3066\u5909\u6570infer\u306b\u8fd4\u3055\u308c\u307e\u3059\u3002 
\u30fb20\u884c\uff1acv2.imencode\u3067bts\u306bnumpy\u306e\u914d\u5217\u3067\u8fd4\u3055\u308c\u307e\u3059\u3002\u5148\u307b\u3069\u306e2x2\u306e\u767d\u753b\u50cf(png)\u306e\u5834\u5408\u306f<\/p>\n\n\n\n
[137  80  78  71  13  10  26  10   0   0   0  13  73  72  68  82   0   0\n   0   2   0   0   0   2   8   2   0   0   0 253 212 154 115   0   0   0\n  22  73  68  65  84   8  29  99 252 255 255  63   3   3   3 227 255 255\n 255  25  24  24   0  50 240   5 253 192  69 117 218   0   0   0   0  73\n  69  78  68 174  66  96 130]<\/code><\/pre>\n\n\n\n
\u30fb21\u884c\uff1atobytes()\u95a2\u6570\u3067byte\u5217\u306b\u5909\u63db\u3057\u307e\u3059\u3002<\/p>\n\n\n\n
b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIHDR\\x00\\x00\\x00\\x02\\x00\\x00\\x00\\x02\\x08\\x02\\x00\\x00\\x00\\xfd\\xd4\\x9as\\x00\\x00\\x00\\x16IDAT\\x08\\x1dc\\xfc\\xff\\xff?\\x03\\x03\\x03\\xe3\\xff\\xff\\xff\\x19\\x18\\x18\\x002\\xf0\\x05\\xfd\\xc0Eu\\xda\\x00\\x00\\x00\\x00IEND\\xaeB`\\x82'<\/code><\/pre>\n\n\n\n
\u30fb21\u884c\uff1a\u8907\u6570\u753b\u50cf\u304c\u3042\u308b\u5834\u5408\u306f\u7e70\u308a\u8fd4\u3057\u3066\u3001imgbytes\u306b\u914d\u5217\u3067\u4fdd\u7ba1\u3057\u307e\u3059\u3002 
\u30fb23\u884c\uff1ainfer(key=tf.constant('uniq_dummy_id_123'), image_bytes=tf.constant(imgbytes)) \u30ad\u30fc\u3068\u30a4\u30e1\u30fc\u30b8\u30c7\u30fc\u30bf\u3092\u6e21\u3059\u3068\u63a8\u8ad6\u7d50\u679c\u304c\u5f97\u3089\u308c\u307e\u3059\u3002<\/p>\n\n\n\n
\u4f8b\uff11\uff09\u9752\u7cfb\u306e\u30c9\u30e9\u30b4\u30f3\u753b\u50cf[Fig7]\u3068\u8d64\u7cfb\u306e\u30c9\u30e9\u30b4\u30f3\u753b\u50cf[Fig8]\u306e\u30a4\u30e1\u30fc\u30b8\u30ab\u30e9\u30fc\u306e\u63a8\u8ad6\"\"<\/p>\n\n\n\n
\n
c:\\apps\\ai>python tfcolor2.py blue-doragon.png red-doragon.png\n\n['serving_default']\n{'labels': <tf.Tensor 'Tile:0' shape=(None, 14) dtype=string>,\n'scores': <tf.Tensor 'scores:0' shape=(None, 14) dtype=float32>,\n'key': <tf.Tensor 'Identity:0' shape=(None,) dtype=string>}\n{\n'labels': <tf.Tensor: shape=(2, 14), dtype=string, numpy=\narray([[b'orange', b'yellow', b'purplish_red', b'green', b'gray',\nb'yellow_green', b'purple', b'pink', b'red', b'black', b'white',\nb'blue_green', b'blue', b'brown'],\n[b'orange', b'yellow', b'purplish_red', b'green', b'gray',\nb'yellow_green', b'purple', b'pink', b'red', b'black', b'white',\nb'blue_green', b'blue', b'brown']], dtype=object)>,\n'scores': <tf.Tensor: shape=(2, 14), dtype=float32, numpy=\narray([[0.05143434, 0.0465448 , 0.3964859 , 0.0465448 , 0.05680693,\n0.04539329, 0.07087733, 0.05016822, 0.04893166, 0.05823032,\n0.04003182, 0.5781289 , 0.7745404 , 0.2492084 ],\n[0.03903484, 0.04772406, 0.67595094, 0.0924189 , 0.04539329,\n0.05968711, 0.39022017, 0.10650936, 0.07087733, 0.05405774,\n0.04539329, 0.08602025, 0.230078 , 0.60351413]], dtype=float32)>,\n'key': <tf.Tensor: shape=(), dtype=string, numpy=b'uniq_dummy_id_123'>}<\/pre>\n<\/div>\n\n\n\n
\u30fb23\u884c\uff1a\uff12\u753b\u50cf\u306e\u30c7\u30fc\u30bf\u3092\u6e21\u3057\u3066\u3044\u308b\u306e\u3067\uff12\u3064\u306e\u7d50\u679c\u304c\u914d\u5217\u3067\u304b\u3048\u3063\u3066\u304d\u307e\u3059\u3002
\u30fb29\uff5e34\u884c\uff1alabels\u9806\u3068score\u9806\u306f\u5bfe\u5fdc\u3057\u3066\u3044\u307e\u3059\u306e\u3067\u3001\u898b\u3084\u3059\u3044\u3088\u3046\u306b\u3053\u3053\u304b\u3089TOP3\u3092\u53d6\u5f97\u3057\u307e\u3059\u3002-scores\u3068\u3059\u308b\u3068\u964d\u9806\u3067\u30bd\u30fc\u30c8\u3057\u307e\u3059\u3002<\/p>\n\n\n\n
\u30a4\u30e1\u30fc\u30b8\u304b\u3089\u62bd\u51fa\u7d50\u679c\uff1a<\/p>\n\n\n\n
blue-doragon.png
\n1: 0.774540 : blue<\/span>
\n2: 0.578129 : blue_green<\/span>
\n3: 0.396486 : purplish_red<\/span>
\nred-doragon.png
\n1: 0.675951 : purplish_red<\/span>
\n2: 0.603514 : brown<\/span>
\n3: 0.390220 : purple<\/span>
\n<\/pre><\/div>\n\n\n\n
\u203b\u7d50\u679c\u306e\u8868\u793a\u306f\u898b\u3084\u3059\u3044\u3088\u3046\u306b\u9069\u5b9c\u6539\u884c\u304a\u3088\u3073\u8272\u3092\u4ed8\u3051\u3066\u3044\u307e\u3059\u3002<\/span>

GCP Auto ML\u4e0a\u3067\u63a8\u8ad6\u3057\u305f\u7d50\u679c\uff1a[Fig9]
 \"\"<\/p>\n\n\n\n
\u4f3c\u305f\u3088\u3046\u306a\u7d50\u679c\u3067\u3059\u3002Auto ML\u4e0a\u3067\u306f\u8272\u306fTOP2\u307e\u3067\u8868\u793a\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n
\u4f8b\uff12\uff09\u30a4\u30c1\u30b4\u306e\u932f\u8996\u753b\u50cf[Fig10] \u3082\u63a8\u8ad6\u3057\u3066\u307f\u307e\u3057\u3087\u3046\u3002\u4eba\u306b\u306f\u8d64\u304f\u898b\u3048\u307e\u3059\u304cAI\u3067\u306f\u305d\u3053\u307e\u3067\u5206\u304b\u308a\u307e\u305b\u3093\u3002
 \"\" 
Fig10.\u3000\u51fa\u5178\uff1a\u300c\u8d64\u304f\u898b\u3048\u308b\u3044\u3061\u3054\u300dhttp:\/\/www.psy.ritsumei.ac.jp\/~akitaoka\/color16.html<\/a>\u3088\u308a<\/p>\n\n\n\n
\u3053\u306e\u753b\u50cf\u306b\u306f\u8d64\u7cfb\u306e\u8272\u306f\u4f7f\u308f\u308c\u3066\u3044\u307e\u305b\u3093\u304c\u4eba\u9593\u306b\u306f\u932f\u8996\u52b9\u679c\u3067\u8d64\u304f\u898b\u3048\u3066\u3044\u307e\u3059\u3002\u3053\u306e\u3088\u3046\u306a\u932f\u8996\u52b9\u679c\u306f\u4eca\u56de\u306e\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3067\u306f\u5168\u304f\u914d\u616e\u3055\u308c\u3066\u3044\u307e\u305b\u3093\u306e\u3067\u5f53\u7136\u3001\u9752\u3001\u7dd1\u7cfb\u306e\u7d50\u679c\u304c\u3067\u307e\u3059\u3002 <\/p>\n\n\n\n
c:\\apps\\ai>python tfcolor2.py ichigo-red-or-green.jpg \n\n['serving_default']\n{'key': <tf.Tensor 'Identity:0' shape=(None,) dtype=string>,\n 'scores': <tf.Tensor 'scores:0' shape=(None, 14) dtype=float32>,\n 'labels': <tf.Tensor 'Tile:0' shape=(None, 14) dtype=string>}\n{\n'key': <tf.Tensor: shape=(), dtype=string, numpy=b'uniq_dummy_id_123'>,\n'scores': <tf.Tensor: shape=(1, 14), dtype=float32, numpy=\narray([[0.05541629, 0.05968711, 0.06916751, 0.37779844, 0.03903484,\n0.07087733, 0.05823032, 0.04893166, 0.0465448 , 0.05968711,\n0.04893166, 0.85638064, 0.6816758 , 0.09464505]], dtype=float32)>,\n'labels': <tf.Tensor: shape=(1, 14), dtype=string, numpy=>\narray([[b'orange', b'yellow', b'purplish_red', b'green', b'gray', \nb'yellow_green', b'purple', b'pink', b'red', b'black', b'white',b'blue_green', \nb'blue', b'brown']], dtype=object)>}\n
ichigo-red-or-green.jpg\n1: 0.856381 : blue_green<\/span>
\n2: 0.681676 : blue<\/span>
\n3: 0.377798 : green<\/span>
\n<\/pre>\n<\/div>\n\n\n\n
GCP Auto ML\u4e0a\u3067\u63a8\u8ad6\u3057\u305f\u7d50\u679c\u3082\u8f09\u305b\u3066\u304a\u304d\u307e\u3059\u3002[Fig11]
\"\"<\/p>\n\n\n\n
AI\u3068\u3044\u3048\u3069\u3082\u5b66\u7fd2\u30e2\u30c7\u30eb\u3042\u308a\u304d\u3067\u3059\u3002\u932f\u8996\u306e\u30d1\u30bf\u30fc\u30f3\u3082\u591a\u6570\u7528\u610f\u3057\u3066\u5b66\u7fd2\u3059\u308c\u3070\u5224\u5225\u3067\u304d\u308b\u304b\u3082\u3057\u308c\u307e\u305b\u3093\u3002<\/p>\n\n\n
\u6700\u5f8c\u306b<\/h2>\n\n\n
\u30af\u30e9\u30a6\u30c9\u5074\u3067\u6642\u9593\u306e\u304b\u304b\u308b\u5b66\u7fd2\u3092\u884c\u3063\u3066\u5f97\u3089\u308c\u305f\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3092\u7aef\u672b\u5074\u306b\u4e00\u5ea6\u914d\u4fe1\u3057\u3066\u304a\u3051\u3070\u3001\u7aef\u672b\u5074\u3067\u30af\u30e9\u30a6\u30c9\u3078\u306e\u554f\u3044\u5408\u308f\u305b\u306b\u3088\u308b\u6642\u9593\u30ed\u30b9\u3082\u306a\u304f\u63a8\u8ad6\u304c\u5373\u5ea7\u306b\u3067\u304d\u307e\u3059\u3002\u3053\u308c\u304c\u30a8\u30c3\u30b8AI\u306e\u826f\u3044\u3068\u3053\u308d\u3067\u3059\u3002<\/p>\n\n\n\n
GCP Auto ML\u3067\u5b66\u7fd2\u3057\u305f\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3092\u5229\u7528\u3055\u308c\u305f\u3044\u65b9\u306e\u53c2\u8003\u306b\u306a\u308c\u3070\u3068\u601d\u3044\u307e\u3059\u3002<\/p>\n\n\n\n
\u63a8\u8ad6\u90e8\u5206\u3092\u30d7\u30ed\u30b0\u30e9\u30df\u30f3\u30b0\u3059\u308b\u3068\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3001\u8a55\u4fa1\u3082\u30af\u30e9\u30a6\u30c9\u3067\u306f\u306a\u304f\u3066\u81ea\u5206\u3067\u5b9f\u65bd\u3057\u3066\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3092\u4f5c\u6210\u3057\u305f\u304f\u306a\u3063\u3066\u304f\u308b\u3068\u601d\u3044\u307e\u3059\u3002TensorFlow, Keras,VGG16\u3092\u5229\u7528\u3057\u305f\u30de\u30eb\u30c1\u30e9\u30d9\u30eb\u5206\u985e\u3082\u7b46\u8005\u3067\u306f\u4f5c\u6210\u3057\u3066\u304a\u308a\u307e\u3059\u304c\u3001\u307e\u3060\u691c\u8a3c\u6bb5\u968e\u3067\u3059\u3002\u3044\u305a\u308c\u304b\u306e\u6a5f\u4f1a\u306b\u516c\u958b\u3057\u305f\u3044\u3068\u601d\u3044\u307e\u3059\u3002\uff08\u306a\u304a\u3053\u306e\u5834\u5408\u306f\u8a08\u7b97\u306b\u6642\u9593\u304c\u304b\u304b\u308b\u306e\u3067NIVIDA\u7cfbGPU\u304c\u5fc5\u9808\u3067\u3059\uff09<\/p>\n\n\n
\u25bd\u30ad\u30e3\u30f3\u30da\u30fc\u30f3\u958b\u50ac\u4e2d\uff01<\/strong><\/span><\/span><\/span><\/div>\n
\u30b3\u30a2\u30b5\u30fc\u30d0\u30fc\u3067\u306f\u3001V2\u30d7\u30e9\u30f3\uff08CORE-X,Y,Z\uff09\u3068\u30c9\u30e1\u30a4\u30f3\u306e\u540c\u6642\u7533\u3057\u8fbc\u307f\u3067
\u30c9\u30e1\u30a4\u30f3\u304c<\/span>\u5b9f\u8cea0<\/span>\u5186\uff08\u5e74\u9593\u6700\u59273,889<\/span>\u5186\u304a\u5f97\uff09<\/strong><\/span>\u306b\u306a\u308b\u30b5\u30fc\u30d0\u30fc\u30bb\u30c3\u30c8\u5272\u7279\u5178<\/strong><\/span><\/span>\u3092\u5c55\u958b\u4e2d\u3067\u3059\u3002
<\/span><\/span><\/span><\/div>\n
\u662f\u975e\u3001\u304a\u5f97\u306a\u3053\u306e\u6a5f\u4f1a\u306b\u3054\u5229\u7528\u304f\u3060\u3055\u3044\uff01
\u6700\u65b0\u306e\u30ad\u30e3\u30f3\u30da\u30fc\u30f3\u306f<\/span><\/span>\u3053\u3061\u3089<\/strong><\/a><\/span>\u304b\u3089
<\/span><\/div>\n
<\/p>\n
\"\u8d85\u9ad8\u901f\u5316\u3092\u5b9f\u73fe\u3059\u308b\u30ec\u30f3\u30bf\u30eb\u30b5\u30fc\u30d0\u30fc<\/a><\/p>\n
<\/p>\n\n\n
\u203b\u30e6\u30fc\u30b6\u30fc\u30ce\u30fc\u30c8<\/a>\u306e\u8a18\u4e8b\u306f\u3001\u5f0a\u793e\u30b5\u30fc\u30d3\u30b9\u3092\u3054\u5229\u7528\u306e\u304a\u5ba2\u69d8\u306b\u57f7\u7b46\u3044\u305f\u3060\u3044\u3066\u304a\u308a\u307e\u3059\u3002<\/p>\n
\u57f7\u7b46\u8005\uff1a\u82d7\u5834 \u7fd4\u69d8<\/span><\/div>\n
\n
\u533b\u7642\u30e1\u30fc\u30ab\u30fc\u3067\u65b0\u7d20\u6750\u7814\u7a76\u958b\u767a\u5f8c\u3001\u96fb\u6a5f\u30e1\u30fc\u30ab\u30fc\u3067\u5236\u5fa1\u5668\u7cfb\u30b7\u30b9\u30c6\u30e0\u958b\u767a\u3092\u7d4c\u3066IT\u7cfb\u30de\u30eb\u30c1\u30a8\u30f3\u30b8\u30cb\u30a2\u3092\u3057\u3066\u3044\u307e\u3059\u3002\u307e\u305f\u30c7\u30b6\u30a4\u30f3\u601d\u8003\u3092\u5b9f\u8df5\u3057\u3001\u30a2\u30fc\u30c8\u601d\u8003\u306a\u3069\u306e\u3044\u308d\u3093\u306a\u601d\u8003\u65b9\u6cd5\u306b\u8208\u5473\u304c\u3042\u308a\u307e\u3059\u3002<\/p><\/div>\n","protected":false},"excerpt":{"rendered":"
\u306f\u3058\u3081\u306b \u524d\u56de\u3001Google AutoML Vision\u3092\u4f7f\u3063\u3066\u300cAI\u753b\u50cf\u89e3\u6790\u3067\u30a4\u30e1\u30fc\u30b8\u30ab\u30e9\u30fc\u3092\u62bd\u51fa\u3059\u308b\u300d\u3092\u7d39\u4ecb\u3057\u307e\u3057\u305f\u3002(\u624b\u9806\u8a73\u7d30\u306f\u3053\u3061\u3089) Google\u306eAutoML Vision\u306f\u3042\u3089\u304b\u3058\u3081\u7528\u610f\u3055\u308c\u3066\u3044\u308b\u5b66\u7fd2\u6e08\u307f\u306e […]<\/p>\n","protected":false},"author":1,"featured_media":2544,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[15],"tags":[],"acf":[],"_links":{"self":[{"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/posts\/2529"}],"collection":[{"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/comments?post=2529"}],"version-history":[{"count":5,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/posts\/2529\/revisions"}],"predecessor-version":[{"id":2545,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/posts\/2529\/revisions\/2545"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/media\/2544"}],"wp:attachment":[{"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/media?parent=2529"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/categories?post=2529"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/media.value-domain.com\/wp-json\/wp\/v2\/tags?post=2529"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}