Today we're sharing our first step towards spatial intelligence: an AI system that generates 3D worlds from a single image.
èæ¯ ã¡ã¿ãã¼ã¹ãARçã®é²å±ã§äººä½ã®3Dã¢ãã«åã¯éè¦å¤§ ï¼Vtuberçã®ã¢ãã¿ã¼,ã²ã¼ã ã¢ã¼ã·ã§ã³ä½æãæ åã³ã³ãã³ãã®ä½æçï¼ ä»åã¯ç»åãåç»ããã©ããã£ã¦äººä½ã3Dåãããã¨ããæè¡ãç´¹ä»ãããã¨æãã¾ãã 3Dã¢ãã«ã®è¡¨ç¾æ¹æ³ ç»åãã3Dã¢ãã«ãä½æããä¸ã§éè¦ãªã®ã¯ï¼Dãã©ã®ãããªå½¢ã§è¡¨ç¾ãããã°ããNNã®å¦ç¿ã«é©ãã¦ãããã¨ãããã¨ã§ãã 3Dã¢ãã«ãPoint Cloudã®ãããªç¹ç¾¤ã¨ãã¦è¡¨ç¾ããã®ããã¡ãã·ã¥ã¨ãã¦è¡¨ç¾ããã®ããªã©åã3Dã«ãã¦ãå¤æ°ã®è¡¨ç¾æ¹æ³ãããã¾ããç¨éãNNã®å¦ç¿ã«é©ãã表ç¾å½¢å¼ ãå¤æ°ææ¡ããã¦ãããä»åã¯SMPLã¨NeRFã¨ããï¼ã¤ã®è¡¨ç¾æ¹æ³ã«é¢ãã¦ãç´¹ä»ãããã¾ãã æ¼ããã¦ããããè¦ç´ æè¡1: SMPLã¢ãã« SMPLã¨ã¯? ãã©ã¡ã¼ã¿åããã人ä½ã®3Dã¢ãã« SMPL: A Skinned Multi-Person Linear M
ããã«ã¡ã¯ãStableDiffusion2.0çºè¡¨ã®éã«ãç»åã®æ·±åº¦æ å ±ãå ã«è¢«åä½ã®å½¢ç¶ãæãªããã¨ãªãç»åçæãè¡ãDepth to Image Diffusion Modelãå ¬éããã¦ãã¾ãããã試ãã¦ã¿ã¾ããã¨ããçµæ§åãã£ãã®ã§ãç´¹ä»ãã¾ãã æ©ã触ãã¦ãï¼ã¨è¨ãæ¹ã¯huggingfaceã®ãã¢çãä¸çªãæ軽ãµã¯ãµã¯ã«è©¦ããã¨æãã¾ãã®ã§ã©ããã https://huggingface.co/spaces/radames/stable-diffusion-depth2img ã¾ãhuggingfaceã®ãã¢ã§ã¯è§£å度ã512*512ã§åºå®ããã¦ãã¾ãããcolabçã§ã¯é«è§£å度çæãå¯è½ã§ããããã¨githubã¯ããã ã¡ãªã¿ã«ãã£ã¨è¦ãéãã§ã¯Automatic1111ãªã©ã®web uiã«ã¯ä»ãã¨ãã¾ã å®è£ ããã¦ãªãã¿ããã§ãããæ¤ç´¢ããã¨DepthMapMaskã¨ãmul
ããã©ã¤ãã®äºæ¬¡åµä½ã«ã¤ã㦠ã¿ãªãã¾ã«æããããã£ã©ã¯ã¿ã¼ãçã¿åºãã¦ããéç¨ã«ã¯ãæºããå ¨ã¦ã®æ¹ã ã®å§åçãªæãããã£ã¦ãã¾ãã â ã¯ãªã¨ã¤ã¿ã¼ã®ã¿ãªãã¾ã«ããã¦ããããã©ã¤ãã®äºæ¬¡åµä½ã³ã³ãã³ããå¶ä½ããã ããã¨ã§æ´»åãæ¯ãã¦ããã ãã ã²ãã¦ã¯å½ãè¶ãã¦æ¡æ£ãèªç¥ãããã¾ã§ã®çºå±ã«å¯ä¸ãã¦ãã ãã£ããã¨ãæ¹ãã¦å¾¡ç¤¼ç³ãä¸ãã¾ãã â ä»å¾ã¨ãããå¤ãã®ã¿ãªãã¾ã«å ã ã¨å®å¿ãã¦åµä½æ´»åãè¡ã£ã¦ããã ãããã« æ¹ãã¦ãäºæ¬¡çåµä½ã©ã¤ã»ã³ã¹è¦ç´ã«ã¤ãã¦ãç解ã¨ãååããé¡ããã¾ãã â âo ããã©ã¤ãäºæ¬¡çåµä½ã©ã¤ã»ã³ã¹è¦ç´ ã¿ãªãã¾ã¸ã®ãããã ã©ã¤ã»ã³ã¹è¦ç´ããç解è³ããéµå®ããã ããã¨ã§ã å½ç¤¾ã¯ã¿ãªãã¾ã®äºæ¬¡åµä½æ´»åã«å¯¾ãã¦å¯å®¹ãªå§¿å¢ãåããã¨ãã§ãã¾ãã â äºæ¬¡åµä½ã§ãããããããã«ã¼ã«ãããã¼ãå®ããã¨ã大åã§ãã ã¿ã¬ã³ããå«ãããã¨ãä¸å¿«ã«æããã³ã³ãã³ãã¯èªããã
3ã¤ã®è¦ç¹ âï¸ NeRFã¨ã¯æ°è¦è¦ç¹ã®ç»åçæãããã¯ã¼ã¯ã§ãã。 âï¸Â NeRFã®å ¥åã¯ï½¤5次å ï¼ç©ºé座æ¨ã®x,y,zã¨è¦ç¹ã®Î¸,Ïï¼ã§ï½¤åºåã¯ä½ç©å¯åº¦ï¼âéææï¼ã¨æ¾å°è¼åº¦ï¼âRGBã«ã©ã¼ï¼ã§ãã。 âï¸Â NeRFã«ãã£ã¦å¾æ¥ãããè¤éãªå½¢ç¶ãæã¤å¯¾è±¡ç©ã®æ°è¦è¦ç¹ç»åãå¾ããã¨ã«æåãã。 NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis written by Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng (Submitted on 19 Mar 2020 (v1), last revised 3 Aug 2020 (this version,
Abstract Recent breakthroughs in text-to-image synthesis have been driven by diffusion models trained on billions of image-text pairs. Adapting this approach to 3D synthesis would require large-scale datasets of labeled 3D assets and efficient architectures for denoising 3D data, neither of which currently exist. In this work, we circumvent these limitations by using a pretrained 2D text-to-image
ããã«ã¡ã¯ããããã¼ã§ãã ä»åã¯Three.js(WebGL)ã§ä¸å¹´ä»¥ä¸åå¼·ããææã¨ä¾¿å©ãªã¯ã©ã¹/ã©ã¤ãã©ãªãç´¹ä»ãããã¨æãã¾ãã Three.js(WebGL)ã¨ã¯ï¼ Three.jsã¯ãå°ãã§ãç°¡åã«Webãµã¤ãä¸ã«3Dã®ã³ã³ãã³ãã表示ããã©ã¤ãã©ãªã§ãã ã©ã¤ãã©ãªãªãã§ã«ã¡ã©ã»å½±ã»ã©ã¤ãã»ã¡ãã·ã¥å®è£ ããå ´åã¯ãããªãã®é«åº¦ãªæè¡ãå¿ è¦ã«ãªãJavascriptã®ã³ã¼ãéãè¨å¤§ã«ãªã£ã¦ããã¾ãã Three.jsã®ãããªã©ã¤ãã©ãªãå°å ¥ãããã¨ã§ãJavascriptã®åºç¤ã¨ãThree.jsã®ã¯ã©ã¹ãå©ç¨ãããã¨ã§æ°è»½ã«3Dã®ã³ã³ãã³ãã表示ãããã¨ãã§ãã¾ãã å½åã¯ãThree.jsãåå¼·ããããã©ã¾ãã©ãããæãã¤ããã°ããããããããå ¬å¼ããã¥ã¡ã³ããè¦ãã¨è±æã§ä½ãã©ãããã°ããã®ãããããªããã¨æãã¾ãã Three.jsãããããåå¼·ããã«ã¯ã以ä¸ã®ãµã¤ãã
ã¿ã¤ãã«ã®éããªãã§ãããé©å½ã«æ£æ©ããªãã 3D ã¹ãã£ã³ãããã¾ãã£ã¦ããã®ã 2020 年代æåã«æ¥ãæ°ããªéã³ã ã¨ããçµè«ã«è³ã£ãã®ã§ãã¿ãªããä»ãã iPhone 12 Pro ã Pro Max ãè²·ã£ã¦ããã¾ãããããã£ã¦ãªã人ã¯ã ããç解ãã¦ããªãã¨æãã¾ãããæé«ã«æ¥½ããã§ãã ãããã iPhone 㧠3D ã¹ãã£ã³ã£ã¦ä½ã ãã£ã¦è©±ã§ãããä»å¹´ã§ã iPad Pro ãã LiDAR ãè¼ãããã«ãªã£ã¦ãéã«è¨ãã¨æ®éã«ã«ã¡ã©ã®æ å ±ã«æ·±åº¦ï¼ã«ã¡ã©ããç©ä½ã¸ã®è·é¢ã§ãï¼ãä¹ããããããã«ãªã£ããã§ããããããããã§ãã㨠AR ã§ä½¿ã£ã¦ç¾å®ä¸çã«ããæãã«ç©ãä¹ã£ããã®ãç®ç㧠Apple ã¨ãã¦ãä»ããç©ã£ã½ãããå®éããããã¢ããªããã£ã±ãåºã¦ãã¾ããã§ããæ£ç´ AR ã¨ã Apple Glasses ãåºãã¾ã§ã©ãã§ããããããªãã§ããããããªã£ã¦ãã㨠LiDA
ãã¡ã³ã¯ã©ãã¸ã¢ããã°ã¬ã¼ãããã¨ãã·ã§ããã¸æ»ããã¨ã¯ã§ãã¾ããã ã·ã§ããããã¡ã³ã¯ã©ãã«ã¢ããã°ã¬ã¼ãããã¨ããã¡ã³ã¯ã©ãã®æ©è½ãå©ç¨ã§ããããã«ãªãã¾ãã ã¢ããã°ã¬ã¼ããããã¨ã§ã§ãããã¨ã»ãã©ã³ã®éè¨Â  ãã¡ã³ã¯ã©ãã§ã¯ãç¡æãã©ã³ã®ã»ãããå¸æã®ä¼è²»ã§ãã¡ã³ã®æ¹ã«éå®ç¹å ¸ãæä¾ãããææãã©ã³ããä½æã§ãã¾ããã¾ããååã«å¯¾ãã¦ãã©ã³éå®ã§è²©å£²ãããã¨ãå¯è½ã§ãã ã»æ稿æ©è½Â  誰ã§ãèªç±ã«è¦ããã¨ãåºæ¥ããå ¬éã³ã³ãã³ããã¨ããã¡ã³ã«ãªããªãã¨è¦ããã¨ãã§ããªãããã«è¨å®åºæ¥ããéå®ã³ã³ãã³ãããä½æã§ãã¾ãã ã»ã³ããã·ã§ã³æ©è½Â  ãã¡ã³ã®ãªã¯ã¨ã¹ãã«çãã¦ç´åãããã¨ã§å ±é ¬ãå¾ãããä»çµã¿ã§ããæ¡ä»¶ã«åã£ããªã¯ã¨ã¹ãã®ã¿å¼ãåãããã¨ãã§ããããã空ãæéãå©ç¨ãã¦ç¡çãªãã¯ããããã¾ãã 詳ããã¯ãã¡ããã覧ãã ããã
ããã«ã¡ã¯ã ç§ã¯ç¾å¨ã¯ãªã¨ã¤ãã£ããã£ããããã¼ã¨ãããããã³ãã¨ã³ãï¼WebGL å®è£ ããä»äºã2å¹´åã»ã©ãã£ã¦ãã¾ãã 1å¹´åæ±äº¬ã®ä¼ç¤¾ã§åãããã¨åå¹´éããªã¼ã©ã³ã¹ããã¦ããã®å¾ã¢ã ã¹ãã«ãã ã®ä¼ç¤¾ã«å ¥ç¤¾ãã¦ç¾å¨8ã¶æçµã¡ã¾ãã three.jsã§çµµãä½ãã®ã好ãã§ãä»äºã§ããã©ã¤ãã¼ãã§ããããªãããã®ãã®ãä½ã£ã¦ãã¾ããã»ã¨ãã©twitterã«ããã¦ãã®ã§ãèå³ãããæ¹ã¯ãã²è¦ã¦ã¿ã¦ãã ããã æè¿ãã¤ãã¿ã¼ã®DMã§ã©ããã£ã¦three.jsãåå¼·ããã°ãããã¢ããã¤ã¹ã欲ããã¨ããã®ãããããã¨ãå¤ããªã£ã¦ãããã§ããããã®è³ªåã«çããã®ã¯é£ãããªã¨æãã¦ãã¾ãã 人ã«ãã£ã¦å¾æä¸å¾æãããããå§ãããã¨æã£ãæç¹ã§ã©ã®ãããããã°ã©ãã³ã°ãæ°å¦ã«ç²¾éãã¦ããã人ããããããã¦ãããããã¹ããªåå¼·æ³ï¼ãããããã°èª°ã§ã大ä¸å¤«ã¨ã¯è¨ããªãã§ãã ãã®è¨äºã§ã¯ãç§ãåå¿è ã¬ãã«ã
3Dããªã³ã¿ã¼ã家ã«ããã¨çæ´»ã¯å¤§ããå¤ããã¾ãããã®ã®èãæ¹ãå¤ããã¾ããã¡ãã£ã¨ããä¸ä¾¿ãããã¨ããããã解決ããç©ããã¶ã¤ã³ãã¦åºåããããã¨ãªãã¾ãã ç§ã¯ã¹ããã®ã¢ããªéçºè ã§ãèªåã§ä½¿ãã¢ããªã®å¤ãã¯èªåã§ä½ã£ã¦ä½¿ã£ã¦ãã¾ãããããçæ´»ãã¹ã¦ã«åºãã£ãæãã§ãã便å©ï¼æ¥½ããã ä»åã¯ã3Dããªã³ã¿ã¼ãè²·ã£ã¦CADãå¦ãã§ç«ä½ãæããã¾ä½ãæ¹æ³ã解説ãã¾ããç«ä½ã¯èª°ã§ãä½ãã¾ãããã ããã1é±éãããããã£ã¦ãã°æã£ãå½¢ã®ãã®ãä½ããããã«ãªãã¾ãã 2018å¹´ã®5æã«ãCADã3Dããªã³ã¿ã¼ãã¾ã£ããåãããªãç¶æ ããå§ãã¦3æ¥ç¨åº¦ã§å¤§ä½ç解ã§ãã¾ãããã ãã§ããããããã§ç¿å¾ã§ããã¨æãã¾ãã ä»ã®CADã¯æ¬å½ã«é©ããããç°¡åã«ä½¿ãã¾ããå°å¦çã§ã使ããã¬ãã«ã§ãããé£ãããããªãã¦è¨ã£ã¦ãªãã§è§¦ã£ã¦ã¿ãã°ããã 決ãã¦é£ãããã®ã§ã¯ããã¾ããããããããããªãããã ãã§ãã
ãç±³å½çºããã¸ã§ã¯ããªã®ã«ãæ¯æ´è ã®3å²ãæ¥æ¬äººãã3Dã¢ãã«ãç«ä½è¦ã§ããâéæãªç®±âLooking Glassãæ¥æ¬ä¸é¸ ãç±³å½çºã®ã¯ã©ã¦ããã¡ã³ãã£ã³ã°ãªã®ã«ãæ¯æ´è ã®3å²ãæ¥æ¬äººã ã£ããââæ¥æ¬ã®ã¯ãªã¨ã¤ã¿ã¼ãç±è¦ç·ãéãâéæãªç®±âãæ¥æ¬ã«ä¸é¸ããã ç±³Looking Glass Factoryã¯2æ26æ¥ã3Dãã¼ã¿ãç«ä½è¦ã§ããéæãªç®±å½¢ãã£ã¹ãã¬ã¤ãLooking Glassãã®å注ãã¯ã©ã¦ããã¡ã³ãã£ã³ã°ãµã¤ããMakuakeãã§å§ãããæ¥æ¬èªããã¥ã¢ã«ã¨æ¥æ¬èªå¯¾å¿ã«ã¹ã¿ãã¼ãµãã¼ããä»å±ããæ¥æ¬åãã¢ãã«ã§ãå®æ©ãå ¥æã§ããæ¯æ´é¡ã¯8.9ã¤ã³ãã¢ãã«ã6ä¸4000åï¼ç¨è¾¼ã以ä¸åï¼ã15.6ã¤ã³ãã¢ãã«ã30ä¸5000åããã Looking Glassã¯ãVRï¼ä»®æ³ç¾å®ï¼ãARï¼æ¡å¼µç¾å®ï¼ã´ã¼ã°ã«ãªã©ã使ããã«ã3Dã¢ãã«ãéæãªç®±ã®ä¸ã«ããããã«è¡¨ç¤ºã§ããæ®ãç½®ãå
VRoidããã¸ã§ã¯ãã¯ããåµä½æ´»åããã£ã¨æ¥½ãããªãå ´æãåµãããç念ã¨ãããã¯ã·ãæ ªå¼ä¼ç¤¾ã«ãã3Däºæ¥ã§ãã 誰ããåæ§è±ããªèªåã®3Dãã£ã©ã¯ã¿ã¼ã¢ãã«ãæã¡ããã®ãã£ã©ã¯ã¿ã¼ãåµä½æ´»åãã³ãã¥ãã±ã¼ã·ã§ã³ã«æ´»ç¨ãããã¨ãã§ããã1人1ã¢ãã¿ã¼ãã®ä¸çãç§ãã¡ã®ããã·ã§ã³ã¯ããã®æªæ¥ããã¯ããã¸ã¼ã¨ã¯ãªã¨ã¤ãã£ãã®åã§å®ç¾ãããã¨ã§ãã VRoidã¯ãçµµãæãããã«ãã£ã©ã¯ã¿ã¼ãä½ããã¨ãã§ãã3Dã¢ããªã³ã°ã½ããã¦ã§ã¢ãVRoid Studioãããçã¾ãã¾ãããç¾å¨ã¯ãã®ã½ããã¦ã§ã¢ãã¯ããã¨ããæ軽ã«ã¢ãã¿ã¼ã¥ããã楽ãããã¹ãã¼ããã©ã³ã¢ããªãã3Dã¢ãã«ãæ稿ã§ãããã©ãããã©ã¼ã ãããã¨é£æºããããã®éçºè ããããã¢ãã¿ã¼ãã¡ãã·ã§ã³ãä¸ã®ä¸ã«ææ¡ãããã©ã³ãäºæ¥ãªã©ãå¤æ¹é¢ã§å±éãè¡ã£ã¦ãã¾ãã
ãã©ã¦ã¶ã§WebGLã使ããããã«ãªã£ã¦3DCGããã°ã©ãã³ã°ã¯ããã¶ã身è¿ãªãã®ã«ãªãã¾ãããã¨æ¸ãã¦ããã°ããéåæãæãããããçã®WebGLãJavaScriptã§æ¸ãã®ã¯æ·å± ãé«ãã£ãããã¾ããã§ããªãã¯ãªããã©åæã¨ãªãç¥èãããªãå¿ è¦ãªæãã three.jsãç»å ´ããã¨ãã¯ãããã§æ®éã«3DCGãã§ããã¨ãããã¨ã§ä¸æ°ã«ã²ãã¾ãã¾ãããã¨ã¯ãããããã§ãã¾ã ãããã¨ã¯å¤ããç»é¢ã«åè§ãç®±ã表示ããå ´å以ä¸ã®ãããªããã°ã©ã ãæ¸ããã¨ã«ãªãã¾ãã ã»ã·ã¼ã³ãä½æ ã»ã©ã¤ããä½æãä½ç½®ã¨åããè¨å®ãã·ã¼ã³ã«è¿½å ã»ã«ã¡ã©ãä½æãä½ç½®ã¨åããè¨å®ãã·ã¼ã³ã«è¿½å ã»ãããªã¢ã«ãä½æãè²ãæå® ã»BoxGeometryãä½æããµã¤ãºãæå® ã»ã¡ãã·ã¥ãä½æãä½ç½®ã¨åããè¨å®ãã·ã¼ã³ã«è¿½å ã»ã¬ã³ãã©ã¼ãä½æ ã»ã¬ã³ããªã³ã°ã«ã¼ãå¦ç ãããã®ã²ã¨ã¤ã§ãééããããã©ã¡ã¼ã¿ã¼ãé©åã§ãª
ãªãªã¼ã¹ãé害æ å ±ãªã©ã®ãµã¼ãã¹ã®ãç¥ãã
ææ°ã®äººæ°ã¨ã³ããªã¼ã®é ä¿¡
å¦çãå®è¡ä¸ã§ã
j次ã®ããã¯ãã¼ã¯
kåã®ããã¯ãã¼ã¯
lãã¨ã§èªã
eã³ã¡ã³ãä¸è¦§ãéã
oãã¼ã¸ãéã
{{#tags}}- {{label}}
{{/tags}}