{"id":34668,"date":"2024-12-10T19:57:08","date_gmt":"2024-12-10T11:57:08","guid":{"rendered":"https:\/\/17aitech.com\/?p=34668"},"modified":"2024-12-10T19:57:08","modified_gmt":"2024-12-10T11:57:08","slug":"eccv-2024-%e6%8f%90%e5%8d%87gpt-4v%e3%80%81gemini%e6%a3%80%e6%b5%8b%e4%bb%bb%e5%8a%a1%e6%80%a7%e8%83%bd%ef%bc%8c%e4%bd%a0%e9%9c%80%e8%a6%81%e8%bf%99%e7%a7%8d%e6%8f%90%e7%a4%ba%e8%8c%83%e5%bc%8f","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=34668","title":{"rendered":"ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-07-22-11\" target=\"_blank\">ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f<\/a><\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png\"><\/a><\/section>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"166\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<section>AIxiv\u4e13\u680f\u662f<mark data-type=\"institutions\" data-id=\"9e96434f-5827-41e3-9b3c-36d39bc0d446\">\u673a\u5668\u4e4b\u5fc3<\/mark>\u53d1\u5e03\u5b66\u672f\u3001\u6280\u672f\u5185\u5bb9\u7684\u680f\u76ee\u3002\u8fc7\u53bb\u6570\u5e74\uff0c<mark data-type=\"institutions\" data-id=\"9e96434f-5827-41e3-9b3c-36d39bc0d446\">\u673a\u5668\u4e4b\u5fc3<\/mark>AIxiv\u4e13\u680f\u63a5\u6536\u62a5\u9053\u4e862000\u591a\u7bc7\u5185\u5bb9\uff0c\u8986\u76d6\u5168\u7403\u5404\u5927\u9ad8\u6821\u4e0e\u4f01\u4e1a\u7684\u9876\u7ea7\u5b9e\u9a8c\u5ba4\uff0c\u6709\u6548\u4fc3\u8fdb\u4e86\u5b66\u672f\u4ea4\u6d41\u4e0e\u4f20\u64ad\u3002\u5982\u679c\u60a8\u6709\u4f18\u79c0\u7684\u5de5\u4f5c\u60f3\u8981\u5206\u4eab\uff0c\u6b22\u8fce\u6295\u7a3f\u6216\u8005\u8054\u7cfb\u62a5\u9053\u3002\u6295\u7a3f\u90ae\u7bb1\uff1aliyazhou@jiqizhixin.com\uff1bzhaoyunfeng@jiqizhixin.com<\/section>\n<\/section>\n<\/blockquote>\n<p><strong>\u672c\u6587\u4f5c\u8005\u6765\u81ea\u6d59\u6c5f\u5927\u5b66\u3001\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u3001\u9999\u6e2f\u4e2d\u6587\u5927\u5b66\u3001\u6089\u5c3c\u5927\u5b66\u548c\u725b\u6d25\u5927\u5b66\u3002\u4f5c\u8005\u5217\u8868\uff1a\u5434\u9038\u7487\uff0c\u738b\u9038\u821f\uff0c\u5510\u8bd7\u7fd4\uff0c\u5434\u6587\u704f\uff0c\u8d3a\u901a\uff0cWanli Ouyang\uff0cPhilip Torr\uff0cJian Wu\u3002\u5176\u4e2d\uff0c\u5171\u540c\u7b2c\u4e00\u4f5c\u8005\u5434\u9038\u7487\u662f\u6d59\u6c5f\u5927\u5b66\u535a\u58eb\u751f\uff0c\u738b\u9038\u821f\u662f\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u79d1\u7814\u52a9\u7406\u3002\u901a\u8baf\u4f5c\u8005\u5510\u8bd7\u7fd4\u662f\u9999\u6e2f\u4e2d\u6587\u5927\u5b66\u535a\u58eb\u540e\u7814\u7a76\u5458\u3002<\/strong><\/p>\n<section>\u591a\u6a21\u6001\u5927\u6a21\u578b\uff08Multimodal Large Language Models\uff0cMLLMs\uff09\u5728\u4e0d\u540c\u7684\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u4e86\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684\u80fd\u529b\uff0c\u5c3d\u7ba1\u5982\u6b64\uff0c\u8fd9\u4e9b\u6a21\u578b\u5728\u68c0\u6d4b\u4efb\u52a1\u4e2d\u7684\u6f5c\u529b\u4ecd\u88ab\u4f4e\u4f30\u3002\u5728\u590d\u6742\u7684<mark data-type=\"tech_tasks\" data-id=\"f181e2bb-a801-4bd4-8c02-469e43aa5411\">\u76ee\u6807\u68c0\u6d4b<\/mark>\u4efb\u52a1\u4e2d\u9700\u8981\u7cbe\u786e\u5750\u6807\u65f6\uff0cMLLMs \u5e26\u6709\u7684\u5e7b\u89c9\u53c8\u8ba9\u5b83\u5e38\u5e38\u9519\u8fc7\u76ee\u6807\u7269\u4f53\u6216\u7ed9\u51fa\u4e0d\u51c6\u786e\u7684\u8fb9\u754c\u6846\u3002\u4e3a\u4e86\u8ba9 MLLMs \u8d4b\u80fd\u68c0\u6d4b\uff0c\u73b0\u6709\u7684\u5de5\u4f5c\u4e0d\u4ec5\u9700\u8981\u6536\u96c6\u5927\u91cf\u9ad8\u8d28\u91cf\u7684\u6307\u4ee4\u6570\u636e\u96c6\uff0c\u8fd8\u9700\u8981\u5bf9\u5f00\u6e90\u6a21\u578b\u8fdb\u884c\u5fae\u8c03\u3002\u8d39\u65f6\u8d39\u529b\u7684\u540c\u65f6\uff0c\u4e5f\u65e0\u6cd5\u5229\u7528\u95ed\u6e90\u6a21\u578b\u66f4\u5f3a\u5927\u7684\u89c6\u89c9\u7406\u89e3\u80fd\u529b\u3002\u4e3a\u6b64\uff0c\u6d59\u6c5f\u5927\u5b66\u8054\u5408\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u548c\u725b\u6d25\u5927\u5b66\u63d0\u51fa\u4e86 <strong>DetToolChain<\/strong>\uff0c\u4e00\u79cd\u91ca\u653e\u591a\u6a21\u6001\u5927<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\u68c0\u6d4b\u80fd\u529b\u7684\u65b0\u63d0\u793a\u8303\u5f0f\u3002\u4e0d\u9700\u8981\u8bad\u7ec3\u5c31\u80fd\u8ba9\u591a\u6a21\u6001\u5927\u6a21\u578b\u5b66\u4f1a\u7cbe\u786e\u68c0\u6d4b\u3002\u76f8\u5173\u7814\u7a76\u5df2\u7ecf\u88ab<strong>\u00a0ECCV 2024 \u6536\u5f55<\/strong>\u3002<\/section>\n<section><\/section>\n<section>\u4e3a\u4e86\u89e3\u51b3 MLLM \u5728\u68c0\u6d4b\u4efb\u52a1\u4e0a\u7684\u95ee\u9898\uff0cDetToolChain \u4ece\u4e09\u4e2a\u70b9\u51fa\u53d1\uff1a\uff081\uff09\u9488\u5bf9\u68c0\u6d4b\u8bbe\u8ba1\u89c6\u89c9\u63d0\u793a\uff08visual prompts\uff09\uff0c\u6bd4\u4f20\u7edf\u7684\u6587\u5b57\u63d0\u793a\uff08textual prompts\uff09\u66f4\u76f4\u63a5\u6709\u6548\u7684\u8ba9 MLLM \u7406\u89e3\u4f4d\u7f6e\u4fe1\u606f\uff0c\uff082\uff09\u628a\u7cbe\u7ec6\u7684\u68c0\u6d4b\u4efb\u52a1\u62c6\u89e3\u6210\u5c0f\u800c\u7b80\u5355\u7684\u4efb\u52a1\uff0c\uff083\uff09\u5229\u7528 chain-of-thought \u9010\u6b65\u4f18\u5316\u68c0\u6d4b\u7ed3\u679c\uff0c\u4e5f\u5c3d\u53ef\u80fd\u7684\u907f\u514d\u591a\u6a21\u6001\u5927\u6a21\u578b\u7684\u5e7b\u89c9\u3002\u00a0<\/section>\n<section><\/section>\n<section>\u4e0e\u4e0a\u8ff0\u7684 insights \u5bf9\u5e94\uff0cDetToolChain \u5305\u542b\u4e24\u4e2a\u5173\u952e\u8bbe\u8ba1\uff1a\uff081\uff09\u4e00\u5957\u5168\u9762\u7684\u89c6\u89c9\u5904\u7406\u63d0\u793a\uff08visual processing prompts\uff09\uff0c\u76f4\u63a5\u5728\u56fe\u50cf\u4e2d\u7ed8\u5236\uff0c\u53ef\u4ee5\u663e\u8457\u7f29\u5c0f\u89c6\u89c9\u4fe1\u606f\u548c\u6587\u672c\u4fe1\u606f\u4e4b\u95f4\u7684\u5dee\u8ddd\u3002\uff082\uff09\u4e00\u5957\u5168\u9762\u7684\u68c0\u6d4b\u63a8\u7406\u63d0\u793a (detection reasoning prompts)\uff0c\u589e\u5f3a\u5bf9\u68c0\u6d4b\u76ee\u6807\u7684\u7a7a\u95f4\u7406\u89e3\uff0c\u5e76\u901a\u8fc7\u6837\u672c\u81ea\u9002\u5e94\u7684\u68c0\u6d4b\u5de5\u5177\u94fe\u9010\u6b65\u786e\u5b9a\u6700\u7ec8\u7684\u76ee\u6807\u7cbe\u786e\u4f4d\u7f6e\u3002<\/section>\n<section><\/section>\n<section>\u901a\u8fc7\u5c06 DetToolChain \u4e0e MLLM \u7ed3\u5408\uff0c\u5982 GPT-4V \u548c Gemini\uff0c\u53ef\u4ee5\u5728\u65e0\u9700\u6307\u4ee4\u8c03\u4f18\u7684\u60c5\u51b5\u4e0b\u652f\u6301\u5404\u79cd\u68c0\u6d4b\u4efb\u52a1\uff0c\u5305\u62ec\u5f00\u653e\u8bcd\u6c47\u68c0\u6d4b\u3001\u63cf\u8ff0<mark data-type=\"tech_tasks\" data-id=\"f181e2bb-a801-4bd4-8c02-469e43aa5411\">\u76ee\u6807\u68c0\u6d4b<\/mark>\u3001\u6307\u79f0\u8868\u8fbe\u7406\u89e3\u548c\u5b9a\u5411<mark data-type=\"tech_tasks\" data-id=\"f181e2bb-a801-4bd4-8c02-469e43aa5411\">\u76ee\u6807\u68c0\u6d4b<\/mark>\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-c9cded1978f2b4918bf9be6272a059e2.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-c9cded1978f2b4918bf9be6272a059e2.png\"><\/a><\/p>\n<ul>\n<li>\n<section>\u8bba\u6587\u6807\u9898\uff1aDetToolChain: A New Prompting Paradigm to Unleash Detection Ability of MLLM<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2403.12488<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section><strong>\u4ec0\u4e48\u662f DetToolChain\uff1f<\/strong><\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-93b46e515c62ea8c77040d3c66dc1229.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-93b46e515c62ea8c77040d3c66dc1229.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe 1 DetToolChain \u7684\u6574\u4f53\u6846\u67b6<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u5982\u56fe 1 \u6240\u793a\uff0c\u5bf9\u4e8e\u7ed9\u5b9a\u7684<mark data-type=\"concepts\" data-id=\"bf740558-f0f7-41a8-87a0-e695a97563b3\">\u67e5\u8be2<\/mark>\u56fe\u50cf\uff0cMLLM \u88ab\u6307\u793a\u8fdb\u884c\u4ee5\u4e0b\u6b65\u9aa4\uff1a<\/section>\n<section><\/section>\n<section>I. Formatting\uff1a\u5c06\u4efb\u52a1\u7684\u539f\u59cb\u8f93\u5165\u683c\u5f0f\u8f6c\u5316\u4e3a\u9002\u5f53\u7684\u6307\u4ee4\u6a21\u677f\uff0c\u4f5c\u4e3a MLLM \u7684\u8f93\u5165\uff1b<\/section>\n<section>II. Think\uff1a\u5c06\u7279\u5b9a\u7684\u590d\u6742\u68c0\u6d4b\u4efb\u52a1\u5206\u89e3\u4e3a\u66f4\u7b80\u5355\u7684\u5b50\u4efb\u52a1\uff0c\u5e76\u4ece\u68c0\u6d4b\u63d0\u793a\u5de5\u5177\u5305\u4e2d\u9009\u62e9\u6709\u6548\u7684\u63d0\u793a\uff08prompts\uff09\uff1b<\/section>\n<section>III. Execute\uff1a\u6309\u987a\u5e8f\u8fed\u4ee3\u6267\u884c\u7279\u5b9a\u7684\u63d0\u793a\uff08prompts\uff09\uff1b<\/section>\n<section>IV. Respond\uff1a\u8fd0\u7528 MLLM \u5176\u81ea\u8eab\u7684\u63a8\u7406\u80fd\u529b\u6765\u76d1\u7763\u6574\u4e2a\u68c0\u6d4b\u8fc7\u7a0b\u5e76\u8fd4\u56de\u6700\u7ec8\u54cd\u5e94\uff08final answer\uff09\u3002<\/section>\n<section><\/section>\n<section><strong>\u68c0\u6d4b\u63d0\u793a\u5de5\u5177\u5305\uff1aVisual Processing Prompts<\/strong><\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-9d9fa5a2a69ca727bc9a4e38635ba38f.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-9d9fa5a2a69ca727bc9a4e38635ba38f.png\"><\/a><\/p>\n<section><em><sup>\u56fe 2\uff1avisual processing prompts \u7684\u793a\u610f\u56fe\u3002\u6211\u4eec\u8bbe\u8ba1\u4e86 (1) Regional Amplifier, (2) Spatial Measurement Standard, \u00a0(3) Scene Image Parser\uff0c\u4ece\u4e0d\u540c\u7684\u89d2\u5ea6\u6765\u63d0\u5347 MLLMs \u7684\u68c0\u6d4b\u80fd\u529b\u3002<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u5982\u56fe 2 \u6240\u793a\uff0c(1) Regional Amplifier \u65e8\u5728\u589e\u5f3a MLLMs \u5bf9\u611f\u5174\u8da3\u533a\u57df\uff08ROI\uff09\u7684\u53ef\u89c1\u6027\uff0c\u5305\u62ec\u5c06\u539f\u59cb\u56fe\u50cf\u88c1\u526a\u6210\u4e0d\u540c\u90e8\u5206\u5b50\u533a\u57df\uff0c\u91cd\u70b9\u5173\u6ce8\u76ee\u6807\u7269\u4f53\u6240\u5728\u5b50\u533a\u57df\uff1b\u6b64\u5916\uff0c\u653e\u5927\u529f\u80fd\u5219\u4f7f\u5f97\u53ef\u4ee5\u5bf9\u56fe\u50cf\u4e2d\u7279\u5b9a\u5b50\u533a\u57df\u8fdb\u884c\u7ec6\u7c92\u5ea6\u89c2\u5bdf\u3002<\/section>\n<section><\/section>\n<section>(2) Spatial Measurement Standard \u901a\u8fc7\u5728\u539f\u59cb\u56fe\u50cf\u4e0a\u53e0\u52a0\u5e26\u6709\u7ebf\u6027\u523b\u5ea6\u7684\u5c3a\u5b50\u548c\u6307\u5357\u9488\uff0c\u4e3a<mark data-type=\"tech_tasks\" data-id=\"f181e2bb-a801-4bd4-8c02-469e43aa5411\">\u76ee\u6807\u68c0\u6d4b<\/mark>\u63d0\u4f9b\u66f4\u660e\u786e\u7684\u53c2\u8003\uff0c\u5982\u56fe 2 (2) \u6240\u793a\u3002\u8f85\u52a9\u5c3a\u5b50\u548c\u6307\u5357\u9488\u4f7f MLLMs \u80fd\u591f\u5229\u7528\u53e0\u52a0\u5728\u56fe\u50cf\u4e0a\u7684\u5e73\u79fb\u548c\u65cb\u8f6c\u53c2\u8003\u8f93\u51fa\u51c6\u786e\u7684\u5750\u6807\u548c\u89d2\u5ea6\u3002\u672c\u8d28\u4e0a\uff0c\u8fd9\u4e00\u8f85\u52a9\u7ebf\u7b80\u5316\u4e86\u68c0\u6d4b\u4efb\u52a1\uff0c\u4f7f MLLMs \u80fd\u591f\u8bfb\u53d6\u7269\u4f53\u7684\u5750\u6807\uff0c\u800c\u4e0d\u662f\u76f4\u63a5\u9884\u6d4b\u5b83\u4eec\u3002<\/section>\n<section><\/section>\n<section>(3) Scene Image Parser \u6807\u8bb0\u9884\u6d4b\u7684\u7269\u4f53\u4f4d\u7f6e\u6216\u5173\u7cfb\uff0c\u5229\u7528\u7a7a\u95f4\u548c\u4e0a\u4e0b\u6587\u4fe1\u606f\u5b9e\u73b0\u5bf9\u56fe\u50cf\u7684\u7a7a\u95f4\u5173\u7cfb\u7406\u89e3\u3002Scene Image Parser \u53ef\u4ee5\u5206\u4e3a\u4e24\u7c7b\uff1a<strong>\u9996\u5148\uff0c\u9488\u5bf9\u5355\u4e2a\u76ee\u6807\u7269\u4f53<\/strong>\uff0c\u6211\u4eec\u7528\u8d28\u5fc3\u3001\u51f8\u5305\u548c\u5e26\u6807\u7b7e\u540d\u79f0\u53ca\u6846\u7d22\u5f15\u7684\u8fb9\u754c\u6846\u6807\u8bb0\u9884\u6d4b\u7684\u7269\u4f53\u3002\u8fd9\u4e9b\u6807\u8bb0\u4ee5\u4e0d\u540c\u683c\u5f0f\u8868\u793a\u7269\u4f53\u4f4d\u7f6e\u4fe1\u606f\uff0c\u4f7f MLLM \u80fd\u591f\u68c0\u6d4b\u4e0d\u540c\u5f62\u72b6\u548c\u80cc\u666f\u7684\u591a\u6837\u7269\u4f53\uff0c\u7279\u522b\u662f\u5f62\u72b6\u4e0d\u89c4\u5219\u6216\u88ab\u5927\u91cf\u906e\u6321\u7684\u7269\u4f53\u3002\u4f8b\u5982\uff0c\u51f8\u5305\u6807\u8bb0\u5668\u6807\u8bb0\u7269\u4f53\u7684\u8fb9\u754c\u70b9\u5e76\u5c06\u5176\u8fde\u63a5\u4e3a\u51f8\u5305\uff0c\u4ee5\u589e\u5f3a\u5bf9\u5f62\u72b6\u975e\u5e38\u4e0d\u89c4\u5219\u7684\u7269\u4f53\u7684\u68c0\u6d4b\u6027\u80fd\u3002<strong>\u5176\u6b21\uff0c\u9488\u5bf9\u591a\u76ee\u6807<\/strong>\uff0c\u6211\u4eec\u901a\u8fc7\u573a\u666f\u56fe\u6807\u8bb0\u5668\uff08scene graph marker\uff09\u8fde\u63a5\u4e0d\u540c\u7269\u4f53\u7684\u4e2d\u5fc3\uff0c\u4ee5\u7a81\u51fa\u56fe\u50cf\u4e2d\u7269\u4f53\u4e4b\u95f4\u7684\u5173\u7cfb\u3002\u57fa\u4e8e\u573a\u666f\u56fe\uff0cMLLM \u53ef\u4ee5\u5229\u7528\u5176\u4e0a\u4e0b\u6587\u63a8\u7406\u80fd\u529b\u6765\u4f18\u5316\u9884\u6d4b\u7684\u8fb9\u754c\u6846\u5e76\u907f\u514d\u5e7b\u89c9\u3002\u4f8b\u5982\uff0c\u5982\u56fe 2 (3) \u6240\u793a\uff0cJerry \u8981\u5403\u5976\u916a\uff0c\u56e0\u6b64\u5b83\u4eec\u7684 bounding box \u5e94\u8be5\u975e\u5e38\u63a5\u8fd1\u3002<\/section>\n<section><\/section>\n<section><strong>\u68c0\u6d4b\u63d0\u793a\u5de5\u5177\u5305\uff1aDetection Reasoning Prompts<\/strong><\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-705a0688e5bb37239e8a576d464d226d.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-705a0688e5bb37239e8a576d464d226d.png\"><\/a><\/p>\n<section>\u4e3a\u4e86\u63d0\u9ad8\u9884\u6d4b\u6846\u7684\u53ef\u9760\u6027\uff0c\u6211\u4eec\u8fdb\u884c\u4e86\u68c0\u6d4b\u63a8\u7406\u63d0\u793a\uff08\u5982\u8868 1 \u6240\u793a\uff09\uff0c\u4ee5\u68c0\u67e5\u9884\u6d4b\u7ed3\u679c\u5e76\u8bca\u65ad\u53ef\u80fd\u5b58\u5728\u7684\u6f5c\u5728\u95ee\u9898\u3002\u9996\u5148\uff0c\u6211\u4eec\u63d0\u51fa\u4e86 Problem Insight Guider\uff0c\u7a81\u51fa\u56f0\u96be\u95ee\u9898\u5e76\u4e3a<mark data-type=\"concepts\" data-id=\"bf740558-f0f7-41a8-87a0-e695a97563b3\">\u67e5\u8be2<\/mark>\u56fe\u50cf\u63d0\u4f9b\u6709\u6548\u7684\u68c0\u6d4b\u5efa\u8bae\u548c\u76f8\u4f3c\u4f8b\u5b50\u3002\u4f8b\u5982\uff0c\u9488\u5bf9\u56fe 3\uff0cProblem Insight Guider \u5c06\u8be5<mark data-type=\"concepts\" data-id=\"bf740558-f0f7-41a8-87a0-e695a97563b3\">\u67e5\u8be2<\/mark>\u5b9a\u4e49\u4e3a\u5c0f\u7269\u4f53\u68c0\u6d4b\u7684\u95ee\u9898\uff0c\u5e76\u5efa\u8bae\u901a\u8fc7\u653e\u5927\u51b2\u6d6a\u677f\u533a\u57df\u6765\u89e3\u51b3\u5b83\u3002\u5176\u6b21\uff0c\u4e3a\u4e86\u5229\u7528 MLLMs \u56fa\u6709\u7684\u7a7a\u95f4\u548c\u4e0a\u4e0b\u6587\u80fd\u529b\uff0c\u6211\u4eec\u8bbe\u8ba1\u4e86 Spatial Relationship Explorer \u548c Contextual Object Predictor\uff0c\u4ee5\u786e\u4fdd\u68c0\u6d4b\u7ed3\u679c\u7b26\u5408\u5e38\u8bc6\u3002\u5982\u56fe 3 \u6240\u793a\uff0c\u51b2\u6d6a\u677f\u53ef\u80fd\u4e0e\u6d77\u6d0b\u5171\u73b0\uff08\u4e0a\u4e0b\u6587\u77e5\u8bc6\uff09\uff0c\u800c\u51b2\u6d6a\u8005\u7684\u811a\u9644\u8fd1\u5e94\u8be5\u6709\u4e00\u4e2a\u51b2\u6d6a\u677f\uff08\u7a7a\u95f4\u77e5\u8bc6\uff09\u3002\u6b64\u5916\uff0c\u6211\u4eec\u5e94\u7528 Self-Verification Promoter \u6765\u589e\u5f3a\u591a\u8f6e\u54cd\u5e94\u7684\u4e00\u81f4\u6027\u3002\u4e3a\u4e86\u8fdb\u4e00\u6b65\u63d0\u5347 MLLMs \u7684\u63a8\u7406\u80fd\u529b\uff0c\u6211\u4eec\u91c7\u7528\u4e86\u5e7f\u6cdb\u5e94\u7528\u7684 prompting \u65b9\u6cd5\uff0c\u4f8b\u5982 debating \u548c self-debugging \u7b49\u3002\u8be6\u7ec6\u63cf\u8ff0\u8bf7\u89c1\u539f\u6587\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-1d2b6303a70236d5315a231cafcf9a71.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-1d2b6303a70236d5315a231cafcf9a71.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe 3 \u68c0\u6d4b\u63a8\u7406\u63d0\u793a\u53ef\u4ee5\u5e2e\u52a9 MLLMs \u89e3\u51b3\u5c0f\u7269\u4f53\u68c0\u6d4b\u95ee\u9898\uff0c\u4f8b\u5982\uff0c\u5229\u7528\u5e38\u8bc6\u5b9a\u4f4d\u4eba\u811a\u4e0b\u7684\u51b2\u6d6a\u677f\uff0c\u5e76\u9f13\u52b1\u6a21\u578b\u5728\u6d77\u6d0b\u4e2d\u68c0\u6d4b\u51b2\u6d6a\u677f\u3002<\/sup><\/em><\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-21a975d9990a58f1acfce615c8a80dd1.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-21a975d9990a58f1acfce615c8a80dd1.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe 4 \u00a0\u4e00\u4e2a DetToolChain \u5e94\u7528\u4e8e\u65cb\u8f6c<mark data-type=\"tech_tasks\" data-id=\"f181e2bb-a801-4bd4-8c02-469e43aa5411\">\u76ee\u6807\u68c0\u6d4b<\/mark>\u7684\u4f8b\u5b50\uff08HRSC2016 \u6570\u636e\u96c6\uff09<\/sup><\/em><\/section>\n<section><\/section>\n<section><strong>\u5b9e\u9a8c\uff1a\u514d\u8bad\u7ec3\u4e5f\u80fd\u8d85\u8d8a\u5fae\u8c03\u65b9\u6cd5<\/strong><\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-f1891340034fe2663dfc3a569f2fbeb3.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-f1891340034fe2663dfc3a569f2fbeb3.png\"><\/a><\/p>\n<section>\u5982\u8868 2 \u6240\u793a\uff0c\u6211\u4eec\u5728 open vocabulary detection\uff08OVD\uff09\u4e0a\u8bc4\u4f30\u4e86\u6211\u4eec\u7684\u65b9\u6cd5\uff0c\u6d4b\u8bd5\u4e86\u5728 COCO OVD benchmark \u4e2d 17 \u4e2a\u65b0\u7c7b\u300148 \u4e2a\u57fa\u7840\u7c7b\u548c\u6240\u6709\u7c7b\u7684 AP50 \u7ed3\u679c\u3002\u7ed3\u679c\u663e\u793a\uff0c\u4f7f\u7528\u6211\u4eec\u7684 DetToolChain\uff0cGPT-4V \u548c Gemini \u7684\u6027\u80fd\u5747\u663e\u8457\u63d0\u5347\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-667ecf7f2b6cb47256ad5d11cdd33010.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-667ecf7f2b6cb47256ad5d11cdd33010.png\"><\/a><\/p>\n<section>\u4e3a\u4e86\u5c55\u793a\u6211\u4eec\u7684\u65b9\u6cd5\u5728\u6307\u79f0\u8868\u8fbe\u7406\u89e3\u4e0a\u7684\u6709\u6548\u6027\uff0c\u6211\u4eec\u5c06\u6211\u4eec\u7684\u65b9\u6cd5\u4e0e\u5176\u4ed6\u96f6\u6837\u672c\u65b9\u6cd5\u5728 RefCOCO\u3001RefCOCO + \u548c RefCOCOg \u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u4e86\u6bd4\u8f83\uff08\u8868 5\uff09\u3002\u5728 RefCOCO \u4e0a\uff0cDetToolChain \u4f7f\u5f97 GPT-4V \u57fa\u7ebf\u5728 val\u3001test-A \u548c test-B \u4e0a\u7684\u6027\u80fd\u5206\u522b\u63d0\u5347\u4e86 44.53%\u300146.11% \u548c 24.85%\uff0c\u5c55\u793a\u4e86 DetToolChain \u5728 zero-shot \u6761\u4ef6\u4e0b\u4f18\u8d8a\u7684\u6307\u79f0\u8868\u8fbe\u7406\u89e3\u548c\u5b9a\u4f4d\u6027\u80fd\u3002<\/section>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-07-22-11\" target=\"_blank\">ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:ECCV 2024 | [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[27],"tags":[73,68],"class_list":["post-34668","post","type-post","status-publish","format-standard","hentry","category-news","tag-73","tag-68"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f - \u4e00\u8d77AI\u6280\u672f<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/17aitech.com\/?p=34668\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=34668\",\"url\":\"https:\/\/17aitech.com\/?p=34668\",\"name\":\"ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=34668#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=34668#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png\",\"datePublished\":\"2024-12-10T11:57:08+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=34668#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=34668\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=34668#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=34668#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/17aitech.com\/#website\",\"url\":\"https:\/\/17aitech.com\/\",\"name\":\"\u4e00\u8d77AI\u6280\u672f\",\"description\":\"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca\",\"alternateName\":\"\u4e00\u8d77AI\u6280\u672f\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/17aitech.com\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\",\"name\":\"AI\u5c0f\u52a9\u624b\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/image\/\",\"url\":\"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png\",\"contentUrl\":\"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png\",\"caption\":\"AI\u5c0f\u52a9\u624b\"},\"description\":\"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e\",\"url\":\"https:\/\/17aitech.com\/?page_id=33738&user=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f - \u4e00\u8d77AI\u6280\u672f","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/17aitech.com\/?p=34668","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=34668","url":"https:\/\/17aitech.com\/?p=34668","name":"ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=34668#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=34668#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png","datePublished":"2024-12-10T11:57:08+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=34668#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=34668"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=34668#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-43807c090ff78fd394a64b8011117829.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=34668#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"ECCV 2024 | \u63d0\u5347GPT-4V\u3001Gemini\u68c0\u6d4b\u4efb\u52a1\u6027\u80fd\uff0c\u4f60\u9700\u8981\u8fd9\u79cd\u63d0\u793a\u8303\u5f0f"}]},{"@type":"WebSite","@id":"https:\/\/17aitech.com\/#website","url":"https:\/\/17aitech.com\/","name":"\u4e00\u8d77AI\u6280\u672f","description":"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca","alternateName":"\u4e00\u8d77AI\u6280\u672f","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/17aitech.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d","name":"AI\u5c0f\u52a9\u624b","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/#\/schema\/person\/image\/","url":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","contentUrl":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","caption":"AI\u5c0f\u52a9\u624b"},"description":"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e","url":"https:\/\/17aitech.com\/?page_id=33738&user=3"}]}},"_links":{"self":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/34668","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=34668"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/34668\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=34668"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=34668"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=34668"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}