å·çï¼éåå´ ååã®è¨äº(ãæè¡è§£èª¬ãä¼¼ã¦ããæååããããï¼ã¬ã¼ãã³ã·ã¥ã¿ã¤ã³è·é¢ã¨ã¸ã£ãã»ã¦ã£ã³ã¯ã©ã¼è·é¢ã®è¨ç®æ¹æ³ã¨ã¯)ã§ã¯ï¼æååå士ã®é¡ä¼¼åº¦(è·é¢)ãè¨ç®ã§ããææ³ãç´¹ä»ããï¼ã¾ãï¼ãã®è¨äºã®ä¸ã§ï¼èªç¶è¨èªå¦çåéã§ã¯ä¸»ã«ææ¸ï¼æååï¼éåçã«ã¤ãã¦é¡ä¼¼åº¦ãè¨ç®ããå ´é¢ãå¤ããã¨ã«ã¤ãã¦ã触ããï¼ä»åã¯éåå士ã®é¡ä¼¼åº¦ã表ç¾ãã以ä¸ã®ï¼ã¤ã®ä¿æ°ã¨è¨ç®æ¹æ³ã«ã¤ãã¦è§£èª¬ããï¼ âJaccardä¿æ° âDiceä¿æ° âSimpsonä¿æ° ãã®åã«ï¼èªç¶è¨èªå¦çã§é¡ä¼¼åº¦ã表ãææ¨ã«ã¤ãã¦ç¢ºèªãããï¼ èªç¶è¨èªå¦çã§ä½¿ç¨ãããé¡ä¼¼åº¦(è·é¢) èªç¶è¨èªå¦çã®åéã§ã¯ï¼é¡ä¼¼åº¦ã測ã対象ã«ãã£ã¦ææ³ã使ãåããï¼ ããã§ã¯äºåã«ï¼ä¸»ã«ä½¿ç¨ãããææ³ã«ã¤ãã¦ç¢ºèªãã¦ãããï¼ ãã¯ãã«å士ã®é¡ä¼¼åº¦ ãâã³ãµã¤ã³é¡ä¼¼åº¦ âãã¢ã½ã³ã®ç¸é¢ä¿æ° âåå·®ãã¿ã¼ã³é¡ä¼¼åº¦ éåå士ã®é¡ä¼¼åº¦(ä»åã®è§£èª¬å¯¾è±¡) ã
çµæã¯åºã¦ããããä½ãããããæ°ãããã»ã»ã»ãããªæã®ãã§ãã¯ãªã¹ã ãã¼ã¿ãæºåãã¦ããã¼ã«ã«æå ¥ãã¦å®è¡ããããäºæ³ãã¦ããã®ã¨ã¯å¤§åéãçµæãåºã¦ãã¦æ¸æãããªãã¦ãã¨ã¯æ¥å¸¸è¶é£¯äºã ãããã®éã©ããã£ã¦æ¤è¨¼ããããããã«ã¤ãã¦ã¾ã¨ããã å ¥åãã¼ã¿ãééãã¦ãã æ£ãããã¼ã¿ã使ã£ã¦ããã¤ããã ã£ãã®ã«èª¿ã¹ã¦ã¿ããééã£ã¦ããã¨ããã®ãæãå¤ãã ããããã®å 容ãæ§ã ã§ããã ãã§ãã¯æ¼ã ãã¼ã¿ãåãåã£ãããã§ãã¯ãããã¨ã¯å¿ è¦ã ããããã§ããã§ãã¯ããæ¼ãã¦ãããããããªãã®ã§æ¹ãã¦ç¢ºèªããããã§ãã¯é ç®ã¯ æ¬ æå¤ ç°å¸¸å¤ å é ã®0ãæ¶ãã¦ãã æ¡ããããã æåã³ã¼ã éè¤ ãã¼ã¿ã足ããªãï¼å¤ããã ãªã©ã ãã¤ã®éã«ãã«ãã¼ã¿ã®ä»æ§ãå¤ãã£ã¦ãã ç¸æã®ãªãã©ã·ã¼ãä½ãã¨å®æçã«åãåã£ã¦ããã¯ãã®ãã¼ã¿ãªã®ã«ãªããçªç¶ä»æ§ãå¤ãã£ã¦ãããã¨ãããã®ã§ã調ã¹ã¦ãééãããã
Convolutional neural networks are designed for dense data, but vision data is often sparse (stereo depth, point clouds, pen stroke, etc.). We present a method to handle sparse depth data with optional dense RGB, and accomplish depth completion and semantic segmentation changing only the last layer. Our proposal efficiently learns sparse features without the need of an additional validity mask. We
ãªãªã¼ã¹ãé害æ å ±ãªã©ã®ãµã¼ãã¹ã®ãç¥ãã
ææ°ã®äººæ°ã¨ã³ããªã¼ã®é ä¿¡
j次ã®ããã¯ãã¼ã¯
kåã®ããã¯ãã¼ã¯
lãã¨ã§èªã
eã³ã¡ã³ãä¸è¦§ãéã
oãã¼ã¸ãéã
{{#tags}}- {{label}}
{{/tags}}