{"id":32,"date":"2026-05-07T22:25:43","date_gmt":"2026-05-07T13:25:43","guid":{"rendered":"https:\/\/material-ai-lab.com\/?p=32"},"modified":"2026-05-07T22:25:43","modified_gmt":"2026-05-07T13:25:43","slug":"%e4%bb%8a%e3%81%a0%e3%81%8b%e3%82%89%e3%81%93%e3%81%9dvggnet%e3%82%92%e5%ad%a6%e3%81%b6%e7%90%86%e7%94%b1%ef%bd%9c%e5%8f%a4%e5%85%b8%e7%9a%84cnn%e3%81%8c%e4%bb%8a%e3%82%82%e9%87%8d%e8%a6%81%e3%81%aa","status":"publish","type":"post","link":"https:\/\/material-ai-lab.com\/?p=32","title":{"rendered":"\u4eca\u3060\u304b\u3089\u3053\u305dVGGNet\u3092\u5b66\u3076\u7406\u7531\uff5c\u53e4\u5178\u7684CNN\u304c\u4eca\u3082\u91cd\u8981\u306a\u308f\u3051"},"content":{"rendered":"\n<p>\u6700\u65b0\u306e\u753b\u50cf\u5206\u985eAI\u306f\u3001<strong>CNN\uff08Convolutional Neural Network\u3001\u7573\u307f\u8fbc\u307f\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\uff09\u7cfb<\/strong>\u3068\u3001<strong>ViT\uff08Vision Transformer\uff09\u7cfb<\/strong>\u306b\u5927\u304d\u304f\u5206\u3051\u3066\u8003\u3048\u3089\u308c\u307e\u3059\u3002CNN\u7cfb\u3067\u306fEfficientNetV2\u3084ConvNeXt\u3001ViT\u7cfb\u3067\u306fVision Transformer\uff08ViT\uff09\u3092\u306f\u3058\u3081\u3068\u3057\u305f\u30e2\u30c7\u30eb\u7fa4\u304c\u5e83\u304f\u77e5\u3089\u308c\u3066\u304a\u308a\u3001\u3069\u3061\u3089\u3082\u9ad8\u3044\u6027\u80fd\u3092\u793a\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3057\u304b\u3057\u3001\u672c\u8a18\u4e8b\u3067\u306f\u3042\u3048\u3066\u6700\u65b0\u30e2\u30c7\u30eb\u306e\u6bd4\u8f03\u305d\u306e\u3082\u306e\u3067\u306f\u306a\u304f\u3001<strong>CNN\u3068\u4e00\u6b21\u8996\u899a\u91ce\u306e\u95a2\u4fc2<\/strong>\u3001\u305d\u3057\u3066\u53e4\u5178\u7684\u306aCNN\u30e2\u30c7\u30eb\u3067\u3042\u308b<strong>VGGNet<\/strong>\u306b\u6ce8\u76ee\u3057\u307e\u3059\u3002\u7406\u7531\u306f\u3001\u30b7\u30f3\u30d7\u30eb\u306aCNN\u306e\u69cb\u9020\u3092\u7406\u89e3\u3059\u308b\u3068\u3001\u753b\u50cf\u8a8d\u8b58\u306e\u57fa\u672c\u3060\u3051\u3067\u306a\u304f\u3001Style Transfer\u3084Perceptual Loss\u306e\u3088\u3046\u306a\u5fdc\u7528\u307e\u3067\u898b\u901a\u3057\u3084\u3059\u304f\u306a\u308b\u304b\u3089\u3067\u3059\u3002VGGNet\u306f\u53e4\u3044\u30e2\u30c7\u30eb\u3067\u3042\u308a\u306a\u304c\u3089\u3001\u4eca\u3067\u3082\u753b\u50cf\u751f\u6210\u3084\u77e5\u899a\u7684\u306a\u985e\u4f3c\u5ea6\u8a55\u4fa1\u306e\u6587\u8108\u3067\u91cd\u8981\u306a\u5f79\u5272\u3092\u6301\u3063\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">1. CNN\u3068\u8996\u899a\u91ce\u306e\u30cb\u30e5\u30fc\u30ed\u30f3\u306b\u3064\u3044\u3066<\/h2>\n\n\n\n<p>CNN\u306e\u7573\u307f\u8fbc\u307f\u5c64\u306f\u3001\u4eba\u9593\u3084\u52d5\u7269\u306e\u8996\u899a\u51e6\u7406\u3068\u5b8c\u5168\u306b\u540c\u3058\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u304c\u3001<strong>\u5c40\u6240\u7684\u306a\u30d1\u30bf\u30fc\u30f3\u3092\u6bb5\u968e\u7684\u306b\u6349\u3048\u308b<\/strong>\u3068\u3044\u3046\u70b9\u3067\u3001\u8996\u899a\u91ce\u306e\u7814\u7a76\u3068\u3057\u3070\u3057\u3070\u95a2\u9023\u3065\u3051\u3066\u8aac\u660e\u3055\u308c\u307e\u3059\u3002\u7279\u306b\u6709\u540d\u306a\u306e\u304c\u3001Hubel \u3068 Wiesel \u306b\u3088\u308b1959\u5e74\u306e\u7814\u7a76\u3067\u3059\u3002\u3053\u306e\u7814\u7a76\u3067\u306f\u3001<strong>\u732b\u306e\u4e00\u6b21\u8996\u899a\u91ce<\/strong>\u306e\u5358\u4e00\u30cb\u30e5\u30fc\u30ed\u30f3\u3092\u8a18\u9332\u3057\u3001\u7279\u5b9a\u306e\u5411\u304d\u306e\u7dda\u3084\u7d30\u9577\u3044\u523a\u6fc0\u306b\u5f37\u304f\u53cd\u5fdc\u3059\u308b\u7d30\u80de\u304c\u3042\u308b\u3053\u3068\u304c\u793a\u3055\u308c\u307e\u3057\u305f\u3002\u3053\u308c\u306f\u3001\u8996\u899a\u91ce\u306b<strong>\u5411\u304d\u9078\u629e\u6027\u3092\u6301\u3064\u53d7\u5bb9\u91ce<\/strong>\u304c\u5b58\u5728\u3059\u308b\u3053\u3068\u3092\u793a\u3057\u305f\u53e4\u5178\u7684\u306a\u6210\u679c\u3067\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEjFSqTRl3ISXk1a5E7idLNF7igPtjB3isNlYuyeK_j93Hel7R4fFHO05Er-_YlspUwU0Ln-NoM8tWBUhZwvMrBVK0xEsxfesylUfcN3GWrpZbetA9tO970esJA1V3bp8yBiv7OUYmlliBSlYov51tlfeP3nxtCslM-l2T4ILHpU8SOz05N_J5xQG8gDC8ka=w640-h637\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Hubel, D. H., &amp; Wiesel, T. N. (1959).&nbsp;<em>Receptive fields of single neurones in the cat&#8217;s striate cortex.<\/em>&nbsp;Journal of Physiology, 148, 574\u2013591<\/p>\n\n\n\n<p>\u3053\u306e\u8003\u3048\u65b9\u306f\u3001CNN\u3092\u7406\u89e3\u3059\u308b\u3046\u3048\u3067\u3082\u76f4\u611f\u7684\u3067\u3059\u3002CNN\u3067\u306f\u3001\u753b\u50cf\u5168\u4f53\u3092\u3044\u304d\u306a\u308a\u7406\u89e3\u3059\u308b\u306e\u3067\u306f\u306a\u304f\u3001\u307e\u305a\u306f<strong>\u5c0f\u3055\u306a\u9818\u57df\u306e\u30a8\u30c3\u30b8\u3001\u5411\u304d\u3001\u8272\u306e\u5909\u5316<\/strong>\u306e\u3088\u3046\u306a\u7279\u5fb4\u3092\u62fe\u3044\u3001\u305d\u306e\u5f8c\u306e\u5c64\u3067\u3088\u308a\u8907\u96d1\u306a\u5f62\u72b6\u3078\u3068\u7d44\u307f\u5408\u308f\u305b\u3066\u3044\u304d\u307e\u3059\u3002\u521d\u5fc3\u8005\u306e\u65b9\u306f\u3001CNN\u3092\u300c\u753b\u50cf\u306e\u4e2d\u304b\u3089\u3001\u7dda\u3084\u6a21\u69d8\u306e\u7279\u5fb4\u3092\u5c11\u3057\u305a\u3064\u898b\u3064\u3051\u3066\u3044\u304f\u4ed5\u7d44\u307f\u300d\u3068\u8003\u3048\u308b\u3068\u5206\u304b\u308a\u3084\u3059\u3044\u3068\u601d\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>CNN\u3067\u3082\u3001\u521d\u671f\u5c64\u306f\u3053\u306e\u3088\u3046\u306a\u5f79\u5272\u3092\u6301\u3064\u3053\u3068\u304c\u77e5\u3089\u308c\u3066\u3044\u307e\u3059\u3002\u305f\u3068\u3048\u3070AlexNet\u306e\u8ad6\u6587\u3067\u306f\u3001\u5b66\u7fd2\u5f8c\u306e\u521d\u671f\u7573\u307f\u8fbc\u307f\u30ab\u30fc\u30cd\u30eb\u304c\u3001<strong>\u7279\u5b9a\u65b9\u5411\u306e\u7dda\u3084\u8272\u306e\u5909\u5316\u306b\u53cd\u5fdc\u3059\u308b\u30d5\u30a3\u30eb\u30bf<\/strong>\u306e\u3088\u3046\u306b\u632f\u308b\u821e\u3046\u3053\u3068\u304c\u78ba\u8a8d\u3067\u304d\u307e\u3059\u3002AlexNet\u81ea\u4f53\u306f\u5927\u304d\u3081\u306e11\u00d711\u7573\u307f\u8fbc\u307f\u3092\u4f7f\u3063\u305f\u30e2\u30c7\u30eb\u3068\u3057\u3066\u6709\u540d\u3067\u3059\u304c\u3001\u305d\u306e\u521d\u671f\u5c64\u304c\u30a8\u30c3\u30b8\u3084\u8272\u306e\u7279\u5fb4\u62bd\u51fa\u306b\u8fd1\u3044\u50cd\u304d\u3092\u3059\u308b\u70b9\u306f\u3001CNN\u306e\u57fa\u672c\u3092\u7406\u89e3\u3059\u308b\u3046\u3048\u3067\u975e\u5e38\u306b\u91cd\u8981\u3067\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEjC0G0C69oThK69CJTjnM5DtnCqz-bJ_5JFn5eaNVWK-a_VtOhRPQNedMY3J5ghoiVMnufTduKy6CEBSYiNrljrNHmw_Ik0X05W4yZpaICrApHSMtcJyO4Q_exJ89zRk9f759yQ0j_u4wL6zc7vUrckGYFBgyvUg-kk7b__1uUxYMdqRiPXG6m9vircW_6-=w400-h347\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Alex Krizhevsky,&nbsp;Ilya Sutskever and Geoffrey E. Hinton (2012),&nbsp;<em>ImageNet Classification with Deep Convolutional Neural Networks<\/em>, Communications of the ACM, 60<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">2. VGGNet\u306e\u7279\u5fb4<\/h2>\n\n\n\n<p>VGGNet\u306f\u3001ILSVRC 2014\u3067\u9ad8\u3044\u6210\u7e3e\u3092\u53ce\u3081\u305f\u4ee3\u8868\u7684\u306a\u753b\u50cf\u5206\u985e\u30e2\u30c7\u30eb\u3067\u3059\u3002\u6700\u5927\u306e\u7279\u5fb4\u306f\u3001<strong>3\u00d73\u306e\u5c0f\u3055\u306a\u7573\u307f\u8fbc\u307f\u3092\u4f55\u5c64\u3082\u7a4d\u307f\u91cd\u306d\u308b<\/strong>\u3068\u3044\u3046\u3001\u975e\u5e38\u306b\u30b7\u30f3\u30d7\u30eb\u306a\u8a2d\u8a08\u306b\u3042\u308a\u307e\u3059\u3002\u8907\u96d1\u306a\u7279\u6b8a\u6a5f\u69cb\u3092\u524d\u9762\u306b\u51fa\u3059\u306e\u3067\u306f\u306a\u304f\u3001\u57fa\u672c\u7684\u306a\u7573\u307f\u8fbc\u307f\u3092\u6df1\u304f\u91cd\u306d\u308b\u3053\u3068\u3067\u6027\u80fd\u3092\u9ad8\u3081\u305f\u70b9\u304c\u3001VGGNet\u306e\u5927\u304d\u306a\u9b45\u529b\u3067\u3059\u3002<\/p>\n\n\n\n<p>\u3053\u306e\u30b7\u30f3\u30d7\u30eb\u3055\u306f\u3001\u521d\u5fc3\u8005\u306b\u3068\u3063\u3066\u3082\u5927\u304d\u306a\u5229\u70b9\u3067\u3059\u3002VGGNet\u306f\u300c\u753b\u50cf\u5206\u985e\u30e2\u30c7\u30eb\u306e\u57fa\u672c\u5f62\u300d\u3092\u5b66\u3076\u306e\u306b\u5411\u3044\u3066\u304a\u308a\u3001<strong>\u7573\u307f\u8fbc\u307f\u3001\u6d3b\u6027\u5316\u95a2\u6570\u3001\u30d7\u30fc\u30ea\u30f3\u30b0\u3001\u5168\u7d50\u5408\u5c64<\/strong>\u3068\u3044\u3063\u305f\u4e3b\u8981\u90e8\u54c1\u304c\u3069\u306e\u3088\u3046\u306b\u9023\u643a\u3059\u308b\u306e\u304b\u3092\u7406\u89e3\u3057\u3084\u3059\u3044\u304b\u3089\u3067\u3059\u3002\u307e\u305f\u3001\u5f8c\u5e74\u306e\u7814\u7a76\u3067\u306f\u3001\u3053\u306e\u7d20\u76f4\u306a\u69cb\u9020\u304c\u753b\u50cf\u306e\u7279\u5fb4\u8868\u73fe\u306b\u3082\u72ec\u7279\u306e\u6271\u3044\u3084\u3059\u3055\u3092\u3082\u305f\u3089\u3057\u3066\u3044\u308b\u3053\u3068\u304c\u5206\u304b\u3063\u3066\u304d\u307e\u3057\u305f\u3002&nbsp;<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEibNvE7G0e-635Jcu-l8rd1pz7pPVDAt7i0pSuxUs1J1zhgyWOkgeNWC5lZK-o9NkFc8Uk-X7dIBeJox6V6O4MMn6hp5VCzWLkioEqQ7cO9GmXsWoxsxrYR7o8wzztdHcFIBuxESakrjBUSS9teyRu9xu2S2TONWg56Le5mfUZhw5uunAucJNapkaoFIsoV=w395-h400\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Simonyan, K., &amp; Zisserman, A. (2015).&nbsp;<em>Very Deep Convolutional Networks for Large-Scale Image Recognition<\/em>. In&nbsp;International Conference on Learning Representations (ICLR 2015)<\/p>\n\n\n\n<p>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">3. VGG\u306e\u7279\u5fb4\u7a7a\u9593\u306fStyle\u3068Content\u3092\u6271\u3044\u3084\u3059\u3044<\/h2>\n\n\n\n<p>VGGNet\u304c\u7279\u306b\u6709\u540d\u306a\u306e\u306f\u3001\u753b\u50cf\u5206\u985e\u6027\u80fd\u3060\u3051\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002Gatys\u3089\u306eNeural Style Transfer\u3067\u306f\u3001VGG\u306e\u4e2d\u9593\u5c64\u306e\u7279\u5fb4\u3092<strong>content\uff08\u5f62\u3084\u914d\u7f6e\u306a\u3069\u306e\u5185\u5bb9\uff09\u306e\u8868\u73fe\u3068\u3057\u3066\u7528\u3044\u3001\u7279\u5fb4\u30de\u30c3\u30d7\u540c\u58eb\u306e\u76f8\u95a2\u3092\u307e\u3068\u3081\u305fGram\u884c\u5217<\/strong>\u3092<strong>style\uff08\u8cea\u611f\u3084\u6a21\u69d8\u3001\u8272\u8abf\u306a\u3069\u306e\u4f5c\u98a8\uff09<\/strong>\u306e\u8868\u73fe\u3068\u3057\u3066\u7528\u3044\u308b\u3053\u3068\u3067\u3001\u4e21\u8005\u3092\u3042\u308b\u7a0b\u5ea6\u5206\u3051\u3066\u6271\u3048\u308b\u3053\u3068\u304c\u793a\u3055\u308c\u307e\u3057\u305f\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEjg_L0bjJL8S7cXYs7SRvrxSLoMoY9e55HgQ6pTwyHne3Ne4kjvhzUQ338fUiz2Fh8xKHWc6SM_6vIvImD3d5OMcDAFcvA9SB2iGXAsgEk8i9N_NcwY2ZBuAwxMjVP_K8HWG7Zxmespo4ZKuIelZAMKYgGqTYzqVe3tyLKnmGpTUEe4tK0k9WgiJZh_Omg9=w451-h640\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Gatys, L. A., Ecker, A. S., &amp; Bethge, M. (2015).&nbsp;<em>A Neural Algorithm of Artistic Style<\/em>. arXiv:1508.06576.<\/p>\n\n\n\n<p>\u3053\u306econtent\u3068style\u3092\u5206\u96e2\u3059\u308b\u80fd\u529b\u306f\u3001<strong>\u3088\u308a\u65b0\u3057\u3044CNN\u7cfb\u5206\u985e\u5668\u304c\u512a\u308c\u308b\u308f\u3051\u3067\u306f\u306a\u304f\u3001\u3080\u3057\u308d\u6a19\u6e96\u7684\u306a VGG \u306e\u307b\u3046\u304c\u5b89\u5b9a\u3057\u3066\u9ad8\u54c1\u8cea\u3067\u3042\u308b<\/strong>\u3053\u3068\u304c\u308f\u304b\u3063\u3066\u3044\u307e\u3059\u3002\u4e0b\u56f3\u306fVGG\u3068ResNet\u3067\u3001\u5b66\u7fd2\u524d(r-)\u3068\u5b66\u7fd2\u5f8c(p-)\u3067Neural Style Transfer\u3092\u884c\u3063\u305f\u7814\u7a76\u3067\u3059\u3002VGG Net\u306f\u5b66\u7fd2\u524d(r-VGG)\u3067\u3082\u3042\u308b\u7a0b\u5ea6\u30b9\u30bf\u30a4\u30eb\u8ee2\u79fb\u3092\u884c\u3048\u3066\u304a\u308a\u3001<strong>VGG\u306e\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u81ea\u4f53\u304ccontent\u3068style\u3092\u5206\u96e2\u3059\u308b\u6a5f\u80fd\u3092\u6709\u3057\u3066\u3044\u308b<\/strong>\u3053\u3068\u304c\u308f\u304b\u3063\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEgvo8SM_957MorczJYEyKObxGgaLdu44bIYmoU_ymn5JzrXy4TEaaI3LrXvLWYglWMQ7ppGVssDJJgeEtWjaTJJB3cDqN7S-yeddhu6xWTquBHLilFCHV62UdvfKpoAT7dYx96RUGqUwBRESRYJfSG9PoJaPw7KDYFeAA018K0H8SEKrplx55eA1cMzQ8Gx=w640-h278\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Wang, P., Li, Y., &amp; Vasconcelos, N. (2021).&nbsp;<em>Rethinking and Improving the Robustness of Image Style Transfer<\/em>. In&nbsp;<em>Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2021)<\/em>.<\/p>\n\n\n\n<p>\u3053\u3053\u3067\u91cd\u8981\u306a\u306e\u306f\u3001\u300cVGG\u304c\u753b\u50cf\u3092\u5358\u306b\u5206\u985e\u3059\u308b\u3060\u3051\u3067\u306a\u304f\u3001<strong>\u4eba\u304c\u898b\u305f\u3068\u304d\u306e\u898b\u305f\u76ee\u306b\u8fd1\u3044\u7279\u5fb4<\/strong>\u3092\u3046\u307e\u304f\u8868\u73fe\u3057\u3084\u3059\u3044\u300d\u3068\u3044\u3046\u70b9\u3067\u3059\u3002\u3082\u3061\u308d\u3093\u3001content\u3068style\u304c\u5b8c\u5168\u306b\u72ec\u7acb\u3057\u3066\u3044\u308b\u308f\u3051\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u304c\u3001VGG\u306e\u7279\u5fb4\u7a7a\u9593\u306f\u3053\u306e2\u3064\u3092\u6bd4\u8f03\u7684\u6271\u3044\u3084\u3059\u304f\u3001Neural Style Transfer\u306e\u6210\u529f\u3092\u652f\u3048\u307e\u3057\u305f\u3002\u521d\u5fc3\u8005\u5411\u3051\u306b\u8a00\u3044\u63db\u3048\u308b\u3068\u3001VGG\u306f\u300c\u4f55\u304c\u63cf\u304b\u308c\u3066\u3044\u308b\u304b\u300d\u3068\u300c\u3069\u3093\u306a\u8cea\u611f\u3084\u96f0\u56f2\u6c17\u304b\u300d\u3092\u5206\u3051\u3066\u8003\u3048\u308b\u571f\u53f0\u3068\u3057\u3066\u4f7f\u3044\u3084\u3059\u3044\u30e2\u30c7\u30eb\u3060\u3068\u8a00\u3048\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">4. Perceptual Loss\u3068\u3057\u3066\u4eca\u3082\u6d3b\u8e8d\u3057\u3066\u3044\u308b<\/h2>\n\n\n\n<p>\u3053\u3046\u3057\u305f\u80cc\u666f\u304b\u3089\u3001VGGNet\u306f**\u300c\u4eba\u9593\u306e\u898b\u305f\u76ee\u306b\u8fd1\u3044\u5dee\u300d\u3092\u6e2c\u308b\u7279\u5fb4\u62bd\u51fa\u5668<strong>\u3068\u3057\u3066\u4eca\u3082\u3088\u304f\u4f7f\u308f\u308c\u307e\u3059\u3002Justin Johnson\u3089\u306f\u3001VGG\u306e\u7279\u5fb4\u7a7a\u9593\u4e0a\u3067\u753b\u50cf\u540c\u58eb\u306e\u5dee\u3092\u6e2c\u308b<\/strong>Perceptual Loss**\u3092\u7528\u3044\u3066\u3001\u30ea\u30a2\u30eb\u30bf\u30a4\u30e0\u306eStyle Transfer\u3084\u8d85\u89e3\u50cf\u3092\u5b9f\u73fe\u3057\u307e\u3057\u305f\u3002\u30d4\u30af\u30bb\u30eb\u5358\u4f4d\u306e\u8aa4\u5dee\u3060\u3051\u3067\u306f\u8868\u73fe\u3057\u306b\u304f\u3044\u300c\u898b\u305f\u76ee\u306e\u81ea\u7136\u3055\u300d\u3092\u3001\u7279\u5fb4\u7a7a\u9593\u306e\u8ddd\u96e2\u3067\u88dc\u3046\u767a\u60f3\u3067\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEgvPgY2J4gVLOmRF3dCkHZaIRUSO9vNW8ufPdYvkpWzivsFKFedsEEj484Mr0Fx1t_uFQ7FsAuiZIgk2w4yOTOYaZeHxmjZ1iXZ8t4-INCepwbnBPHt98Tkp7U5yyhAxP8y2QUIcXoLyzkkMWGvV0BtPCivTwUGttEsusLnFXkgmpjRbk2g_YEySe4XZbPE=w640-h380\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Johnson, J., Alahi, A., &amp; Fei-Fei, L. (2016).&nbsp;<em>Perceptual Losses for Real-Time Style Transfer and Super-Resolution<\/em>. In&nbsp;European Conference on Computer Vision (ECCV 2016)<\/p>\n\n\n\n<p>. arXiv:1603.08155.<\/p>\n\n\n\n<p>\u3053\u306e\u8003\u3048\u65b9\u306f\u3001Style Transfer\u3060\u3051\u306b\u3068\u3069\u307e\u308a\u307e\u305b\u3093\u3002\u753b\u50cf\u751f\u6210\u3084\u753b\u50cf\u5fa9\u5143\u306e\u5206\u91ce\u3067\u306f\u3001<strong>\u307c\u3084\u3051\u305f\u51fa\u529b\u3092\u6e1b\u3089\u3057\u3001\u3088\u308a\u81ea\u7136\u3067\u9ad8\u7cbe\u7d30\u306b\u898b\u3048\u308b\u7d50\u679c\u3092\u5f97\u308b\u305f\u3081<\/strong>\u306b\u3001Perceptual Loss\u304c\u5e83\u304f\u5229\u7528\u3055\u308c\u3066\u3044\u307e\u3059\u3002VAE\uff08Variational Autoencoder\u3001\u5909\u5206\u30aa\u30fc\u30c8\u30a8\u30f3\u30b3\u30fc\u30c0\uff09\u3084\u8d85\u89e3\u50cf\u3001\u753b\u50cf\u518d\u69cb\u6210\u306a\u3069\u3067\u3001\u5358\u7d14\u306a\u753b\u7d20\u8aa4\u5dee\u306b\u52a0\u3048\u3066\u77e5\u899a\u7684\u306a\u640d\u5931\u3092\u5165\u308c\u308b\u8a2d\u8a08\u306f\u4eca\u3067\u3082\u91cd\u8981\u3067\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><a href=\"https:\/\/www.blogger.com\/blog\/post\/edit\/7026973157148678023\/1455086726250227808#\"><img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/a\/AVvXsEhuw_MeYYlI5Pe4ZgCglnRsAd41uxhDjZ5jJ-xuu8zun2v-uXx3MYAq1mpThF3x2x9v5iQyqHkTNSBi3Ke8hTIiUogoKJFm3bJ2qI61z4P564N2zUloHI-Ahj5ZqnUSusywSGx3qOqUBkRQ_6QVl8cnl17RsfCg3Mzdl_7I6wHfU9ZBQenSW-K13Y3gltah=w640-h222\" alt=\"\"\/><\/a><\/figure>\n\n\n\n<p>Hou, X., Shen, L., Sun, K., &amp; Qiu, G. (2016).&nbsp;<em>Deep Feature Consistent Variational Autoencoder<\/em>. arXiv preprint arXiv:1610.00291.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u307e\u3068\u3081<\/h2>\n\n\n\n<p>VGGNet\u306f\u3001\u6700\u65b0\u306e\u753b\u50cf\u5206\u985e\u30e2\u30c7\u30eb\u3068\u6bd4\u3079\u308b\u3068\u53e4\u5178\u7684\u306aCNN\u30e2\u30c7\u30eb\u3067\u3059\u3002\u3057\u304b\u3057\u3001<strong>3\u00d73\u7573\u307f\u8fbc\u307f\u3092\u7a4d\u307f\u91cd\u306d\u308b\u30b7\u30f3\u30d7\u30eb\u306a\u69cb\u9020<\/strong>\u3086\u3048\u306b\u7406\u89e3\u3057\u3084\u3059\u304f\u3001\u3055\u3089\u306b\u753b\u50cf\u306e\u7279\u5fb4\u8868\u73fe\u304c\u6271\u3044\u3084\u3059\u3044\u3068\u3044\u3046\u5f37\u307f\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u307e\u305f\u3001CNN\u306e\u521d\u671f\u5c64\u304c\u30a8\u30c3\u30b8\u3084\u5411\u304d\u306e\u3088\u3046\u306a\u5c40\u6240\u7279\u5fb4\u3092\u6349\u3048\u308b\u3068\u3044\u3046\u8003\u3048\u65b9\u306f\u3001Hubel \u3068 Wiesel \u306e\u53e4\u5178\u7684\u7814\u7a76\u3068\u3082\u91cd\u306d\u3066\u7406\u89e3\u3057\u3084\u3059\u304f\u3001CNN\u306e\u672c\u8cea\u3092\u5b66\u3076\u3046\u3048\u3067\u975e\u5e38\u306b\u6709\u7528\u3067\u3059\u3002VGGNet\u306f\u305d\u306e\u5ef6\u9577\u7dda\u4e0a\u3067\u3001Neural Style Transfer\u3084Perceptual Loss\u3068\u3044\u3063\u305f\u5fdc\u7528\u306b\u4eca\u3082\u6d3b\u7528\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3064\u307e\u308a\u3001VGGNet\u306f\u5358\u306a\u308b\u6614\u306e\u753b\u50cf\u5206\u985e\u30e2\u30c7\u30eb\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002<strong>\u753b\u50cf\u306econtent\u3068style\u3092\u8003\u3048\u308b\u57fa\u76e4<\/strong>\u3068\u3057\u3066\u3001\u305d\u3057\u3066<strong>\u4eba\u9593\u306e\u898b\u305f\u76ee\u306b\u8fd1\u3044\u5dee\u3092\u6271\u3046\u305f\u3081\u306e\u7279\u5fb4\u62bd\u51fa\u5668<\/strong>\u3068\u3057\u3066\u3001\u4eca\u306a\u304a\u5b66\u3076\u4fa1\u5024\u306e\u9ad8\u3044\u30e2\u30c7\u30eb\u3060\u3068\u8a00\u3048\u308b\u3067\u3057\u3087\u3046\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6700\u65b0\u306e\u753b\u50cf\u5206\u985eAI\u306f\u3001CNN\uff08Convolutional Neural Network\u3001\u7573\u307f\u8fbc\u307f\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\uff09\u7cfb\u3068\u3001ViT\uff08Visio&#8230;<\/p>\n","protected":false},"author":1,"featured_media":19,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[4],"tags":[],"class_list":["post-32","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/posts\/32","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=32"}],"version-history":[{"count":1,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/posts\/32\/revisions"}],"predecessor-version":[{"id":33,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/posts\/32\/revisions\/33"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=\/wp\/v2\/media\/19"}],"wp:attachment":[{"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=32"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=32"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/material-ai-lab.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=32"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}