{"id":22792,"date":"2024-10-28T13:58:06","date_gmt":"2024-10-28T05:58:06","guid":{"rendered":"https:\/\/aif.amtbbs.org\/?p=22792"},"modified":"2024-10-28T13:58:06","modified_gmt":"2024-10-28T05:58:06","slug":"%e6%8e%a7%e5%88%b6%e7%94%b5%e8%84%91%e6%89%8b%e6%9c%ba%e7%9a%84%e6%99%ba%e8%83%bd%e4%bd%93%e4%ba%ba%e4%ba%ba%e9%83%bd%e8%83%bd%e9%80%a0%ef%bc%8c%e5%be%ae%e8%bd%af%e5%bc%80%e6%ba%90omniparser","status":"publish","type":"post","link":"https:\/\/aif.amtbbs.org\/index.php\/2024\/10\/28\/22792\/","title":{"rendered":"\u63a7\u5236\u7535\u8111\u624b\u673a\u7684\u667a\u80fd\u4f53\u4eba\u4eba\u90fd\u80fd\u9020\uff0c\u5fae\u8f6f\u5f00\u6e90OmniParser"},"content":{"rendered":"
\u6709\u7f51\u53cb\u53d1\u73b0\u82f9\u679c\u5df2\u7ecf\u9ed8\u9ed8\u53d1\u5e03\u4e86 Ferret-UI \u7684\u4e24\u4e2a\u5b9e\u73b0\u7248\u672c\uff08\u5206\u522b\u57fa\u4e8e Gemma 2B \u548c Llama 8B\uff09\u3002<\/div>\n
\n
\n

\u6700\u8fd1\u8fd9\u51e0\u5929\uff0c\u8ba9\u5927\u6a21\u578b\u5177\u5907\u63a7\u5236\u8ba1\u7b97\u673a\uff08\u5305\u62ec\u7535\u8111\u548c\u624b\u673a\uff09\u7684\u76f8\u5173\u7814\u7a76\u548c\u5e94\u7528\u5982\u96e8\u540e\u6625\u7b0b\u822c\u4e0d\u65ad\u6d8c\u73b0\u3002<\/p>\n

\u5148\u662f Anthropic \u53d1\u5e03\u4e86\u80fd\u63a7\u5236\u8ba1\u7b97\u673a\u7684\u65b0\u7248 Claude 3.5 Sonnet\uff0c\u4e4b\u540e\u8363\u8000 MagicOS 9.0\u00a0\u6765\u4e86\u4e2a\u5168\u5c40\u667a\u80fd\u4f53\uff0c\u518d\u7136\u540e\uff0c\u6628\u5929\u667a\u8c31\u53d1\u5e03\u4e86\u5177\u5907\u300c\u5168\u6808\u5f0f\u5de5\u5177\u4f7f\u7528\u80fd\u529b\u300d\u7684 AutoGLM\uff0c\u540c\u65f6\u534e\u4e3a\u4e5f\u516c\u5e03\u4e86\u4e00\u9879\u53ef\u8ba9 AI \u50cf\u4eba\u7c7b\u4e00\u6837\u64cd\u4f5c\u624b\u673a\u7684\u65b0\u7814\u7a76\u6210\u679c\u00a0LiMAC\u3002<\/p>\n

\u5f88\u663e\u7136\uff0c\u8fd9\u80a1\u70ed\u6f6e\u5b8c\u5168\u6ca1\u6709\u8981\u505c\u606f\u7684\u610f\u601d\u3002\u4eca\u5929\uff0c\u6709\u7f51\u53cb\u53d1\u73b0\u82f9\u679c\u5df2\u7ecf\u9ed8\u9ed8\u53d1\u5e03\u4e86 Ferret-UI \u7684\u4e24\u4e2a\u5b9e\u73b0\u7248\u672c\uff08\u5206\u522b\u57fa\u4e8e Gemma 2B \u548c Llama 8B\uff09\uff0c\u8fd9\u662f\u82f9\u679c\u4eca\u5e74\u4e94\u6708\u53d1\u5e03\u7684\u4e00\u4e2a\u53ef\u8ba9 AI \u7406\u89e3\u624b\u673a\u5c4f\u5e55\u7684\u6280\u672f\uff0c\u8be6\u60c5\u53c2\u9605\u300a\u8ba9\u5927\u6a21\u578b\u7406\u89e3\u624b\u673a\u5c4f\u5e55\uff0c\u82f9\u679c\u591a\u6a21\u6001 Ferret-UI \u7528\u81ea\u7136\u8bed\u8a00\u64cd\u63a7\u624b\u673a\u300b\u3002<\/p>\n

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

\u6765\u81ea X \u7528\u6237\u00a0 Niels Rogge<\/p>\n

    \n
  • Ferret-UI \u9879\u76ee\u5730\u5740\uff1ahttps:\/\/huggingface.co\/papers\/2404.05719<\/li>\n<\/ul>\n

    \u4e0d\u4ec5\u5982\u6b64\uff0c\u5fae\u8f6f\u4e5f\u4f4e\u8c03\u5f00\u6e90\u4e86\u4ed6\u4eec\u7684\u76f8\u5173\u7814\u7a76 OmniParser\uff0c\u8fd9\u662f\u4e00\u4e2a\u57fa\u4e8e\u5927\u6a21\u578b\u7684\u5c4f\u5e55\u89e3\u6790\u5de5\u5177\uff0c\u53ef\u5c06 UI \u622a\u56fe\u8f6c\u6362\u6210\u7ed3\u6784\u5316\u7684\u5143\u7d20\uff1b\u636e\u79f0\u5176\u89e3\u6790\u548c\u7406\u89e3 UI \u7684\u80fd\u529b\u8fbe\u5230\u4e86\u5f53\u524d\u6700\u4f73\u6c34\u5e73\uff0c\u751a\u81f3\u8d85\u8d8a\u4e86 GPT-4V\u3002<\/p>\n

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

      \n
    • \u9879\u76ee\u5730\u5740\uff1ahttps:\/\/huggingface.co\/microsoft\/OmniParser<\/li>\n
    • \u4ee3\u7801\u5730\u5740\uff1ahttps:\/\/github.com\/microsoft\/OmniParser<\/li>\n
    • \u8bba\u6587\u6807\u9898\uff1aOmniParser for Pure Vision Based GUI Agent<\/li>\n
    • \u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/abs\/2408.00203<\/li>\n<\/ul>\n

      \u6709\u4e86\u8fd9\u4e2a\u5de5\u5177\uff0c\u6216\u8bb8\u6bcf\u4e2a\u4eba\u90fd\u53ef\u4ee5\u521b\u5efa\u81ea\u5df1\u7684\u8ba1\u7b97\u673a\u64cd\u63a7\u667a\u80fd\u4f53\u4e86\u3002<\/p>\n

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

      \u5148\u6765\u770b\u770b OmniParser \u7684\u6548\u679c\u3002\u5bf9\u4e8e\u4e00\u4e2a\u7528\u6237\u4efb\u52a1\uff1a\u300c\u5c06\u7ea6\u7ff0\u5185\u65af\u5821\u63d0\u4f9b\u7d20\u98df\u9009\u62e9\u7684\u9910\u5385\u4fdd\u5b58\u5230\u6211\u7684\u884c\u7a0b\u4e2d\u300d\u3002<\/p>\n

      OmniParser \u9996\u5148\u4f1a\u89e3\u6790 Tripadvisor \u7f51\u9875\u5c4f\u5e55\u4e0a\u7684\u6240\u6709\u5143\u7d20\uff0c\u7136\u540e\u5b83\u6210\u529f\u4ece\u4e2d\u627e\u5230\u4e86\u300c\u9910\u5385\u300d\u9009\u9879\u3002\u4e4b\u540e\u5b83\u70b9\u51fb\uff08\u52a8\u4f5c\u6267\u884c\u9700\u8981\u642d\u914d\u5176\u5b83\u6a21\u578b\uff09\u8be5\u9009\u9879\uff0c\u6253\u5f00\u4e86\u4e00\u4e2a\u641c\u7d22\u6846\u3002OmniParser \u7ee7\u7eed\u89e3\u6790\uff0c\u8fd9\u4e00\u6b21\u6ca1\u6709\u5728\u5c4f\u5e55\u4e0a\u627e\u5230\u6240\u9700\u5173\u952e\u8bcd\uff0c\u4e8e\u662f\u5b83\u5728\u641c\u7d22\u6846\u4e2d\u8f93\u5165\u4e86\u300c\u7ea6\u7ff0\u5185\u65af\u5821\u300d\u3002\u518d\u6b21\u89e3\u6790\u540e\uff0c\u5b83\u6253\u5f00\u4e86\u76f8\u5e94\u7684\u641c\u7d22\u9879\uff0c\u5c55\u5f00\u4e86\u641c\u7d22\u7ed3\u679c\u3002\u540c\u6837\uff0c\u7ee7\u7eed\u89e3\u6790\uff0c\u5b83\u6210\u529f\u5b9a\u4f4d\u5230\u4e86\u7d20\u98df\u9009\u9879\uff0c\u7136\u540e\u8fdb\u884c\u4e86\u52fe\u9009\u3002\u6700\u540e\uff0c\u70b9\u51fb\u7b5b\u9009\u51fa\u7684\u7b2c\u4e00\u4e2a\u9009\u9879\u4e0a\u7684\u76f8\u5e94\u6309\u94ae\u5c06\u5176\u6536\u85cf\u5230\u884c\u7a0b\u4e2d\u3002\u81f3\u6b64\uff0c\u4efb\u52a1\u5b8c\u6210\u3002<\/p>\n

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

      \u800c\u5982\u679c\u4f60\u60f3\u770b\u770b\u80fd\u5426\u8fdb\u5165\u5e03\u83b1\u65af\u5ce1\u8c37\u56fd\u5bb6\u516c\u56ed\u5462\uff1fOmniParser \u4e5f\u80fd\u52a9\u4f60\u8f7b\u677e\u5b8c\u6210\u3002<\/p>\n

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

      \u6574\u4f53\u6765\u770b\uff0cOmniParser \u7684\u89e3\u6790\u80fd\u529b\u975e\u5e38\u51fa\u8272\uff0c\u8fc7\u7a0b\u4e5f\u8fd8\u7b97\u6d41\u7545\u3002<\/p>\n

      \u6211\u4eec\u77e5\u9053\uff0c\u4e0d\u540c\u7684\u64cd\u4f5c\u7cfb\u7edf\u548c\u5e94\u7528\u6709\u7740\u5927\u4e0d\u76f8\u540c\u7684\u4ea4\u4e92\u754c\u9762\u3002\u4e3a\u4e86\u7a33\u5065\u5730\u5e94\u5bf9\u5404\u79cd\u4e0d\u540c\u60c5\u51b5\uff0c\u5c4f\u5e55\u89e3\u6790\u6a21\u578b\u9700\u8981\uff1a<\/p>\n

        \n
      1. \u53ef\u9760\u5730\u8bc6\u522b\u4ea4\u4e92\u754c\u9762\u5185\u53ef\u4ea4\u4e92\u7684\u56fe\u6807\uff1b<\/li>\n
      2. \u7406\u89e3\u5c4f\u5e55\u622a\u56fe\u4e2d\u5404\u79cd\u4e0d\u540c\u5143\u7d20\u7684\u542b\u4e49\uff0c\u5e76\u5c06\u8ba1\u5212\u52a8\u4f5c\u4e0e\u5c4f\u5e55\u4e0a\u76f8\u5e94\u7684\u533a\u57df\u51c6\u786e\u5730\u5173\u8054\u8d77\u6765\u3002<\/li>\n<\/ol>\n

        OmniParser \u6b63\u662f\u4e3a\u8fd9\u4e00\u76ee\u7684\u800c\u751f\u7684\u3002OmniParser \u53ef\u57fa\u4e8e\u7528\u6237\u4efb\u52a1\u548c UI \u622a\u56fe\u8f93\u51fa\uff1a(1) \u89e3\u6790\u540e\u7684\u622a\u56fe\uff0c\u5e26\u6709\u8fb9\u754c\u6846\u548c\u6570\u503c ID\uff0c(2) \u5305\u542b\u63d0\u53d6\u51fa\u7684\u6587\u672c\u548c\u56fe\u6807\u63cf\u8ff0\u7684\u5c40\u90e8\u8bed\u4e49\u3002\u4e0b\u9762\u5c55\u793a\u4e86\u51e0\u4e2a\u4f8b\u5b50\uff1a<\/p>\n

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

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

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

        \u5982\u679c\u518d\u642d\u914d\u4e0a\u5176\u5b83\u53ef\u4ee5\u57fa\u4e8e UI \u91c7\u53d6\u884c\u52a8\u7684\u6a21\u578b\uff08\u6bd4\u5982 GPT-4V\u3001Phi-3.5 \u548c Llama 3.2\uff09\uff0c\u4fbf\u53ef\u4ee5\u521b\u9020\u51fa\u53ef\u4ee5\u7406\u89e3\u5e76\u63a7\u5236\u8ba1\u7b97\u673a\u7684\u667a\u80fd\u4f53\u3002<\/p>\n

        \u8be5\u56e2\u961f\u7528 GPT-4V \u505a\u4e86\u5b9e\u9a8c\uff0c\u7ed3\u679c\u53d1\u73b0 OmniParser \u80fd\u5927\u5e45\u63d0\u5347\u5176\u4e3a\u754c\u9762\u533a\u57df\u751f\u6210\u7cbe\u51c6\u52a8\u4f5c\u7684\u80fd\u529b\u3002\u4ed6\u4eec\u4f7f\u7528 OmniParser \u548c GPT-4V \u521b\u5efa\u7684\u4e00\u4e2a\u667a\u80fd\u4f53\u5728 WindowsAgentArena \u57fa\u51c6\u4e0a\u8fbe\u5230\u4e86\u6700\u4f73\u6c34\u5e73\u3002<\/p>\n

        OmniParser \u662f\u5982\u4f55\u70bc\u6210\u7684\uff1f<\/h4>\n

        \u6536\u96c6\u548c\u6574\u7406\u4e13\u7528\u6570\u636e\u96c6 \u2014\u2014OmniParser \u7684\u5f00\u53d1\u59cb\u4e8e\u521b\u5efa\u4e24\u4e2a\u6570\u636e\u96c6\uff1a<\/p>\n

          \n
        • \u4e00\u4e2a\u53ef\u4ea4\u4e92\u533a\u57df\u68c0\u6d4b\u6570\u636e\u96c6\uff0c\u8be5\u6570\u636e\u96c6\u6536\u96c6\u6574\u7406\u81ea\u5e38\u89c1\u7684\u7f51\u9875\uff0c\u5176\u4e2d\u53ef\u70b9\u51fb\u548c\u53ef\u64cd\u4f5c\u7684\u533a\u57df\u90fd\u505a\u4e86\u6807\u6ce8\u3002<\/li>\n
        • \u4e00\u4e2a\u56fe\u6807\u63cf\u8ff0\u6570\u636e\u96c6\uff0c\u65e8\u5728\u5c06\u6bcf\u4e2a UI \u5143\u7d20\u4e0e\u5176\u76f8\u5e94\u7684\u529f\u80fd\u5173\u8054\u8d77\u6765\u3002\u5728\u8bad\u7ec3\u6a21\u578b\u4ee5\u7406\u89e3\u68c0\u6d4b\u5230\u7684\u5143\u7d20\u7684\u8bed\u4e49\u65b9\u9762\uff0c\u6b64\u6570\u636e\u96c6\u975e\u5e38\u5173\u952e\u3002<\/li>\n<\/ul>\n

          \u4e0b\u9762\u5c55\u793a\u4e86\u53ef\u4ea4\u4e92\u533a\u57df\u68c0\u6d4b\u6570\u636e\u96c6\u7684\u4e00\u4e9b\u6837\u672c\u793a\u4f8b\u3002<\/p>\n

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

          \u5bf9\u68c0\u6d4b\u548c\u63cf\u8ff0\u6a21\u578b\u8fdb\u884c\u5fae\u8c03 \u2014\u2014OmniParser \u4f7f\u7528\u4e86\u4e24\u4e2a\u4e92\u8865\u7684\u6a21\u578b\uff1a<\/p>\n

            \n
          • \u4e00\u4e2a\u68c0\u6d4b\u6a21\u578b\uff0c\u5728\u53ef\u4ea4\u4e92\u56fe\u6807\u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u4e86\u5fae\u8c03\uff0c\u5176\u80fd\u53ef\u9760\u5730\u8bc6\u522b\u5c4f\u5e55\u622a\u56fe\u4e2d\u7684\u53ef\u64cd\u4f5c\u533a\u57df\u3002<\/li>\n
          • \u4e00\u4e2a\u63cf\u8ff0\u6a21\u578b\uff0c\u5728\u56fe\u6807\u63cf\u8ff0\u6570\u636e\u96c6\u4e0a\u5b8c\u6210\u4e86\u8bad\u7ec3\uff0c\u5176\u4f5c\u7528\u662f\u63d0\u53d6\u68c0\u6d4b\u5230\u7684\u5143\u7d20\u7684\u529f\u80fd\u8bed\u4e49\uff0c\u4e3a\u9884\u671f\u64cd\u4f5c\u751f\u6210\u51c6\u786e\u7b26\u5408\u4e0a\u4e0b\u6587\u7684\u63cf\u8ff0\u3002<\/li>\n<\/ul>\n

            \u6027\u80fd\u8868\u73b0<\/h4>\n

            \u8be5\u56e2\u961f\u5bf9\u81ea\u5df1\u7684\u65b9\u6cd5\u8fdb\u884c\u4e86\u5b9e\u9a8c\u9a8c\u8bc1\u3002\u7ed3\u679c\u53d1\u73b0\uff0cOmniParser \u53ef\u5927\u5e45\u63d0\u5347 GPT-4V \u5728 ScreenSpot \u57fa\u51c6\u4e0a\u7684\u6027\u80fd\u3002<\/p>\n

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

            \u5728 Mind2Web \u57fa\u51c6\u4e0a\uff0cOmniParser + GPT-4V \u7684\u8868\u73b0\u4e5f\u80dc\u8fc7\u53ef\u4ece HTML \u63d0\u53d6\u989d\u5916\u4fe1\u606f\u7684 GPT-4V\u3002<\/p>\n

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

            \u5728 AITW \u57fa\u51c6\u4e0a\uff0cOmniParser \u7684\u8868\u73b0\u4f18\u4e8e\u4e00\u4e2a\u7ecf\u8fc7\u589e\u5f3a\u7684 GPT-4V\u2014\u2014 \u9644\u5e26\u4e86\u4e00\u4e2a\u4f7f\u7528\u89c6\u56fe\u56fe\u5c42\u8bad\u7ec3\u7684\u4e13\u7528 Android \u56fe\u6807\u68c0\u6d4b\u6a21\u578b\u3002<\/p>\n

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

            \u53e6\u5916\uff0c\u5176\u5728\u65b0\u57fa\u51c6 WindowsAgentArena \u4e0a\u4e5f\u8fbe\u5230\u4e86\u6700\u4f73\u6027\u80fd\u3002<\/p>\n

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

            OmniParser \u53ef\u4f5c\u4e3a\u5f53\u524d\u5404\u79cd\u89c6\u89c9 – \u8bed\u8a00\u6a21\u578b\uff08VLM\uff09\u7684\u63d2\u4ef6\u3002\u4e3a\u4e86\u8fdb\u4e00\u6b65\u6f14\u793a\u8fd9\u4e00\u70b9\uff0c\u8be5\u56e2\u961f\u4e5f\u6d4b\u8bd5\u4e86\u5176\u4e0e Phi-3.5-V \u548c Llama-3.2-V \u7684\u7ec4\u5408\u3002<\/p>\n

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

            \u8be5\u56e2\u961f\u8868\u793a\uff1a\u300c\u6211\u4eec\u5e0c\u671b OmniParser \u53ef\u4ee5\u4f5c\u4e3a\u4e00\u79cd\u901a\u7528\u4e14\u6613\u4e8e\u4f7f\u7528\u7684\u5de5\u5177\uff0c\u5728 PC \u548c\u79fb\u52a8\u5e73\u53f0\u4e0a\u89e3\u6790\u7528\u6237\u7684\u5c4f\u5e55\uff0c\u800c\u65e0\u9700\u4f9d\u8d56 HTML \u548c Android \u4e2d\u7684\u89c6\u56fe\u56fe\u5c42\u7b49\u989d\u5916\u4fe1\u606f\u3002\u300d<\/p>\n

            \u4e0d\u77e5\u9053\u5982\u679c\u5c06 OmniParser \u4e0e\u65b0\u7248 Claude 3.5 Sonnet \u7684 Computer Use \u7ed3\u5408\u8d77\u6765\u4f1a\u662f\u4ec0\u4e48\u6548\u679c\uff1f\u53ef\u80fd\u4f1a\u50cf\u8fd9\u4f4d\u7f51\u53cb\u8bf4\u7684\uff0c\u5f88\u8d5e\u5427\u3002<\/p>\n

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

            \u6587\u7ae0\u6765\u81ea\uff1a51CTO<\/p>\n

             <\/p>\n<\/div>\n<\/div>\n

            <\/div>\n

            <\/g><\/svg><\/i> \"Loading\"<\/p>\n

            <\/div>\n","protected":false},"excerpt":{"rendered":"

            \u6709\u7f51\u53cb\u53d1\u73b0\u82f9\u679c\u5df2\u7ecf\u9ed8\u9ed8\u53d1\u5e03\u4e86 Ferret-UI \u7684\u4e24\u4e2a\u5b9e\u73b0\u7248\u672c\uff08\u5206\u522b\u57fa\u4e8e Gemma 2B \u548c Llama […]<\/p>\n

            <\/div>\n

            <\/g><\/svg><\/i> \"Loading\"<\/p>\n

            <\/div>\n","protected":false},"author":56,"featured_media":22793,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[23,20,80],"tags":[1398],"class_list":["post-22792","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-23","category-20","category-80","tag--ai"],"_links":{"self":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/22792","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/users\/56"}],"replies":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/comments?post=22792"}],"version-history":[{"count":1,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/22792\/revisions"}],"predecessor-version":[{"id":22795,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/22792\/revisions\/22795"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media\/22793"}],"wp:attachment":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media?parent=22792"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/categories?post=22792"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/tags?post=22792"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}