人工知能ã§ãƒ¢ãƒŽã‚¯ãƒå†™çœŸã‚’自動ã§ã‚«ãƒ©ãƒ¼ã«ã™ã‚‹æ–¹æ³• : 船木俊介「ディープビジãƒã‚¹ãƒ•ãƒãƒ³ãƒ†ã‚£ã‚¢ã€ãƒ‡ã‚£ãƒ¼ãƒ—ラーニング（Convolutional Neural Networks（畳ã¿è¾¼ã¿ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ï¼‰ï¼‰ã‚’使ã£ã¦ãƒ¢ãƒŽã‚¯ãƒå†™çœŸã‚’自動ã§ã‚«ãƒ©ãƒ¼åŒ–ã™ã‚‹ã‚ˆã€ã¨ã„ã†é¢ç™½ãã†ãªãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆAutomatic ColorizationãŒã‚ã£ãŸã®ã§è©¦ã—ã¦ã¿ãŸã€‚ Rather than reconstructing the entire color RGB image, I trained models to produce two color channels which I concatenate with the grayscale input channel to produce a YUV image. The Y channel is intensity. This ensures that the intensity of the output will always be the
TensorFlowã§ä¼šè©±AIを作ã£ã¦ã¿ãŸã€‚
{{#tags}}- {{label}}
{{/tags}}