{"id":394,"date":"2024-06-12T15:31:23","date_gmt":"2024-06-12T07:31:23","guid":{"rendered":"https:\/\/aitimes.link\/?p=394"},"modified":"2024-06-12T17:58:19","modified_gmt":"2024-06-12T09:58:19","slug":"stable-diffusion-3-kai-yuan-dao-ji-shi","status":"publish","type":"post","link":"https:\/\/aitimes.link\/index.php\/2024\/06\/12\/stable-diffusion-3-kai-yuan-dao-ji-shi\/","title":{"rendered":"Stable Diffusion 3\u5f00\u6e90\u5012\u8ba1\u65f6"},"content":{"rendered":"\n

\u539f\u521b \u65b0\u667a\u5143
https:\/\/mp.weixin.qq.com\/s\/cDLQds50cX_9Ni_TKA1JaQ<\/a>
AI\u751f\u56fe\u683c\u5c40\u5927\u9707\uff01Stable Diffusion 3\u5f00\u6e90\u5012\u8ba1\u65f6\uff0c2B\u5355\u673a\u53ef\u8dd1\u78be\u538b\u95ed\u6e90Midjourney<\/p>\n\n\n\n

\u3010\u65b0\u667a\u5143\u5bfc\u8bfb\u3011<\/strong>\u91cd\u78c5\u6d88\u606f\uff01Stable Diffusion 3\uff0c\u5927\u6982\u7387\u4f1a\u5728\u660e\u5929\u5f00\u6e90\u3002\u8ddd\u79bb2\u6708SD3\u7684\u6a2a\u7a7a\u51fa\u4e16\uff0c\u5df2\u7ecf\u8fc7\u53bb\u4e864\u4e2a\u6708\u3002\u5982\u679c\u6b64\u4e8b\u4e3a\u771f\uff0c\u751f\u56fe\u5708\u5b50\u7b2c\u4e00\u4e2a\u51fa\u73b0\u5f00\u6e90\u78be\u538b\u95ed\u6e90\u7684\u5947\u666f\uff01\u5f3a\u5927\u7684MMDiT\u5168\u65b0\u67b6\u6784\uff0c\u5c06\u5f7b\u5e95\u6539\u53d8AI\u751f\u56fe\u7684\u683c\u5c40\u3002\u73b0\u5728\uff0c\u5168\u4f53AI\u793e\u533a\u90fd\u5728\u7fd8\u9996\u4ee5\u76fc\u3002<\/p>\n\n\n\n

\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/stabilityai-public-packages.s3.us-west-2.amazonaws.com\/Stable+Diffusion+3+Paper.pdf<\/p>\n\n\n\n

\"\u56fe\u7247\"<\/p>\n\n\n\n

\u7528kimi\u505a\u4e86\u4e0b\u89e3\u8bfb\u5982\u4e0b\uff1a<\/p>\n\n\n\n

\u6587\u4ef61\uff1aEmu\u6a21\u578b\u8bba\u6587 (arXiv:2309.15807)<\/h3>\n\n\n\n
    \n
  • \u6807\u9898<\/strong>: Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack<\/li>\n\n\n\n
  • \u4f5c\u8005<\/strong>: Xiaoliang Dai, Ji Hou, \u7b49\u6765\u81eaMeta\u7684GenAI\u56e2\u961f\u3002<\/li>\n\n\n\n
  • \u6838\u5fc3\u5185\u5bb9<\/strong>:\n
      \n
    • \u63d0\u51fa\u4e86\u4e00\u79cd\u540d\u4e3aEmu\u7684\u56fe\u50cf\u751f\u6210\u6a21\u578b\uff0c\u8be5\u6a21\u578b\u901a\u8fc7\u201c\u8d28\u91cf\u8c03\u6574\u201d\uff08quality-tuning\uff09\u6280\u672f\u663e\u8457\u63d0\u9ad8\u4e86\u751f\u6210\u56fe\u50cf\u7684\u7f8e\u5b66\u8d28\u91cf\u3002<\/li>\n\n\n\n
    • Emu\u6a21\u578b\u57281.1\u4ebf\u56fe\u50cf-\u6587\u672c\u5bf9\u4e0a\u9884\u8bad\u7ec3\uff0c\u7136\u540e\u4ec5\u7528\u51e0\u5343\u5f20\u7cbe\u9009\u7684\u9ad8\u8d28\u91cf\u56fe\u50cf\u8fdb\u884c\u5fae\u8c03\uff0c\u4ee5\u63d0\u9ad8\u89c6\u89c9\u5438\u5f15\u529b\u3002<\/li>\n\n\n\n
    • \u5728\u4e0e\u9884\u8bad\u7ec3\u6a21\u578b\u548cSDXLv1.0\u6a21\u578b\u7684\u6bd4\u8f83\u4e2d\uff0cEmu\u5728\u89c6\u89c9\u5438\u5f15\u529b\u65b9\u9762\u53d6\u5f97\u4e86\u66f4\u9ad8\u7684\u504f\u597d\u7387\u3002<\/li>\n\n\n\n
    • \u8bba\u6587\u8fd8\u5c55\u793a\u4e86\u8d28\u91cf\u8c03\u6574\u5bf9\u4e8e\u5176\u4ed6\u67b6\u6784\uff08\u5982\u50cf\u7d20\u6269\u6563\u548c\u63a9\u853d\u751f\u6210\u53d8\u6362\u5668\u6a21\u578b\uff09\u4e5f\u662f\u6709\u6548\u7684\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n

      \u6587\u4ef62\uff1aStable Diffusion 3\u8bba\u6587 (Stable+Diffusion+3+Paper.pdf)<\/h3>\n\n\n\n
        \n
      • \u6807\u9898<\/strong>: Scaling Rectified Flow Transformers for High-Resolution Image Synthesis<\/li>\n\n\n\n
      • \u4f5c\u8005<\/strong>: Patrick Esser, Sumith Kulal, \u7b49\u6765\u81eaStability AI\u3002<\/li>\n\n\n\n
      • \u6838\u5fc3\u5185\u5bb9<\/strong>:\n
          \n
        • \u4ecb\u7ecd\u4e86\u4e00\u79cd\u65b0\u578b\u7684\u751f\u6210\u6a21\u578b\u2014\u2014Rectified Flow\uff0c\u8be5\u6a21\u578b\u901a\u8fc7\u76f4\u7ebf\u8fde\u63a5\u6570\u636e\u548c\u566a\u58f0\uff0c\u5177\u6709\u66f4\u597d\u7684\u7406\u8bba\u5c5e\u6027\u548c\u6982\u5ff5\u7b80\u5355\u6027\u3002<\/li>\n\n\n\n
        • \u901a\u8fc7\u5927\u89c4\u6a21\u7814\u7a76\uff0c\u5c55\u793a\u4e86\u8fd9\u79cd\u65b9\u6cd5\u5728\u9ad8\u5206\u8fa8\u7387\u6587\u672c\u5230\u56fe\u50cf\u5408\u6210\u65b9\u9762\u7684\u4f18\u8d8a\u6027\u80fd\u3002<\/li>\n\n\n\n
        • \u63d0\u51fa\u4e86\u4e00\u79cd\u65b0\u9896\u7684\u57fa\u4e8eTransformer\u7684\u67b6\u6784\uff0c\u7528\u4e8e\u6587\u672c\u5230\u56fe\u50cf\u7684\u751f\u6210\uff0c\u8be5\u67b6\u6784\u4f7f\u7528\u4e0d\u540c\u7684\u6743\u91cd\u5904\u7406\u56fe\u50cf\u548c\u6587\u672ctoken\uff0c\u5141\u8bb8\u56fe\u50cf\u548c\u6587\u672ctoken\u4e4b\u95f4\u7684\u53cc\u5411\u4fe1\u606f\u6d41\u3002<\/li>\n\n\n\n
        • \u5c55\u793a\u4e86\u8be5\u67b6\u6784\u7684\u53ef\u9884\u6d4b\u6269\u5c55\u8d8b\u52bf\uff0c\u5e76\u4e14\u8f83\u4f4e\u7684\u9a8c\u8bc1\u635f\u5931\u4e0e\u901a\u8fc7\u5404\u79cd\u6307\u6807\u548c\u4eba\u7c7b\u8bc4\u4f30\u6539\u5584\u7684\u6587\u672c\u5230\u56fe\u50cf\u5408\u6210\u5f3a\u76f8\u5173\u3002<\/li>\n\n\n\n
        • \u4ed6\u4eec\u7684\u6700\u5927\u578b\u6a21\u578b\u5728\u5b9a\u91cf\u8bc4\u4f30\u548c\u4eba\u7c7b\u504f\u597d\u8bc4\u7ea7\u65b9\u9762\u8d85\u8d8a\u4e86\u73b0\u6709\u7684\u6700\u5148\u8fdb\u6a21\u578b\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n
          \"\"<\/figure>\n","protected":false},"excerpt":{"rendered":"

          \u539f\u521b \u65b0\u667a\u5143https:\/\/mp.weixin.qq.com\/s\/cDLQds50cX_9Ni_TK […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[],"class_list":["post-394","post","type-post","status-publish","format-standard","hentry","category-ai"],"_links":{"self":[{"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/posts\/394","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/comments?post=394"}],"version-history":[{"count":2,"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/posts\/394\/revisions"}],"predecessor-version":[{"id":397,"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/posts\/394\/revisions\/397"}],"wp:attachment":[{"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/media?parent=394"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/categories?post=394"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aitimes.link\/index.php\/wp-json\/wp\/v2\/tags?post=394"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}