以åã®è¨äºã§ãªã¼ãã¨ã³ã³ã¼ãã«ããç°å¸¸æ¤ç¥ã¯å¤ãï¼ã¨æ¸ãã¦ãã¾ãã¾ãããã æè¿ã¯é²åãã¦ããããã§ãã ä»åãç´¹ä»ããè«æã¯ãæ失é¢æ°ã工夫ãããã¨ã§é常ã®ãªã¼ãã¨ã³ã³ã¼ãããã ç°å¸¸æ¤ç¥è½åãä¸ããææ³ã§ãã â»æ¬ç¨¿ã®å³ã¯è«æï¼Improving Unsupervised Defect Segmentation by Applying Structural Similarity To Autoencodersï¼ããå¼ç¨ãã¦ãã¾ãã è«æã®æ¦è¦ é常ã®ãªã¼ãã¨ã³ã³ã¼ãã«ããç°å¸¸æ¤ç¥ã¯ãå¾®å°ãªç°å¸¸ã¯æããããªãã ããã§ãä¸æã®ç»åã«å¯¾ãå°ããªæ ãç¨æãã¦ãè¼åº¦ãããã³ã³ãã©ã¹ããããæ§é æ å ±ãã®é¡ä¼¼åº¦ãè¨ç®ãã¦ç°å¸¸æ¤ç¥ãè¡ãã æ¬ææ³ã使ããã¨ã§ãé常ã®ãªã¼ãã¨ã³ã³ã¼ããVAEã®ç°å¸¸æ¤ç¥ã¨æ¯ã¹ã¦ãAUCã§å¤§å¹ ãªåä¸ãè¦ãããã ç°å¸¸é¨åã®å¯è¦åã«ã¤ãã¦ããé常ã®ãªã¼ãã¨ã³ã³ã¼ããããåªã
{{#tags}}- {{label}}
{{/tags}}