YOLOv2(Keras / TensorFlow)ã§ãƒ‡ã‚£ãƒ¼ãƒ—ラーニングã«ã‚ˆã‚‹ç”»åƒã®ç‰©ä½“検出を行ㆠ- Qiitaã‚„ã£ãŸã“㨠æµè¡Œã‚Šã®ãƒ‡ã‚£ãƒ¼ãƒ—ラーニングを使ã£ã¦ã€ç”»åƒã®ç‰©ä½“検出を行ã„ã¾ã—ãŸã€‚ 今回ã¯ã€YOLOv2ã¨ã„ã†ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ を使ã£ã¦ç‰©ä½“検出を行ãªã£ã¦ã„ã¾ã™ã€‚ YOLO(You Only Look Once)ã¨ã¯ 畳ã¿è¾¼ã¿ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯(CNN:Convolutional Neural Network)を用ã„ãŸç‰©ä½“検出アルゴリズムã§ã™ã€‚ç¾æ™‚点ã§ã¯v1ã¨v2ãŒå˜åœ¨ã—ã¾ã™ã€‚ YOLO(YOLOv1) è«–æ–‡ã¯ã“ã¡ã‚‰(2015å¹´)。 従æ¥ã¨ã¯ç•°ãªã‚Šã€ç”»åƒã‚’ãƒã‚¦ãƒ³ãƒ‡ã‚£ãƒ³ã‚°ãƒœãƒƒã‚¯ã‚¹ã§åˆ†å‰²ã—ã¦ã‚¯ãƒ©ã‚¹åˆ†é¡žã‚’è¡Œãªã£ã¦ã„る。 we frame object detection as a regression problem to spatially separated bounding boxes and associated class probabilities. çµæžœã¨ã—ã¦ã€45FPSã®å‡¦ç†é€Ÿåº¦ã‚’実ç¾ã—
{{#tags}}- {{label}}
{{/tags}}