{"id":1073853,"date":"2024-06-19T14:35:38","date_gmt":"2024-06-19T21:35:38","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-blog-post&#038;p=1073853"},"modified":"2024-11-21T02:04:00","modified_gmt":"2024-11-21T10:04:00","slug":"new-arrival-in-research-12","status":"publish","type":"msr-blog-post","link":"https:\/\/www.microsoft.com\/en-us\/research\/articles\/new-arrival-in-research-12\/","title":{"rendered":"CVPR\u4e0a\u65b0 | \u4ece\u65b0\u89c6\u89d2\u5408\u6210\u3001\u89c6\u9891\u7f16\u89e3\u7801\u5668\u3001\u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\uff0c\u5230\u6587\u672c\u5e03\u5c40\u5206\u6790\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7cbe\u9009\u8bba\u6587"},"content":{"rendered":"\n<p>\u7f16\u8005\u6309\uff1a\u6b22\u8fce\u9605\u8bfb\u201c\u79d1\u7814\u4e0a\u65b0\u201d\u680f\u76ee\uff01\u201c\u79d1\u7814\u4e0a\u65b0\u201d\u6c47\u805a\u4e86\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u6700\u65b0\u7684\u521b\u65b0\u6210\u679c\u4e0e\u79d1\u7814\u52a8\u6001\u3002\u5728\u8fd9\u91cc\uff0c\u4f60\u53ef\u4ee5\u5feb\u901f\u6d4f\u89c8\u7814\u7a76\u9662\u7684\u4eae\u70b9\u8d44\u8baf\uff0c\u4fdd\u6301\u5bf9\u524d\u6cbf\u9886\u57df\u7684\u654f\u9510\u55c5\u89c9\uff0c\u540c\u65f6\u4e5f\u80fd\u627e\u5230\u5148\u8fdb\u5b9e\u7528\u7684\u5f00\u6e90\u5de5\u5177\u3002<\/p>\n\n\n\n<p>\u672c\u5468\uff0c\u4eba\u5de5\u667a\u80fd\u9886\u57df\u6700\u5177\u5b66\u672f\u5f71\u54cd\u529b\u7684\u9876\u7ea7\u4f1a\u8bae\u4e4b\u4e00\uff0cCVPR \u5927\u4f1a\u5728\u7f8e\u56fd\u897f\u96c5\u56fe\u4e3e\u529e\u3002\u56e0\u6b64\uff0c\u8fd9\u671f\u7684\u201c\u79d1\u7814\u4e0a\u65b0\u201d\u4e3a\u5927\u5bb6\u5e26\u6765\u4e86\u56db\u7bc7\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5165\u9009 CVPR 2024 \u7684\u7cbe\u9009\u8bba\u6587\u89e3\u8bfb\uff0c\u6d89\u53ca\u9886\u57df\u6db5\u76d6\u65b0\u89c6\u89d2\u56fe\u50cf\u5408\u6210\u30013D \u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\u3001\u89c6\u9891\u7f16\u89e3\u7801\u5668\u3001\u6587\u672c\u5e03\u5c40\u5206\u6790\u7b49\u591a\u4e2a\u76f8\u5173\u4e3b\u9898\u3002<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"coponerf-\u7edf\u4e00\u5bf9\u5e94\u70b9\u4f30\u8ba1-\u76f8\u673a\u59ff\u6001\u4f30\u8ba1\u548c\u795e\u7ecf\u8f90\u5c04\u573a\u91cd\u5efa-\u5b9e\u73b0\u7aef\u5230\u7aef\u53cc\u89c6\u56fe\u65b0\u89c6\u89d2\u5408\u6210-cvpr-2024-highlight\u8bba\u6587\">CoPoNeRF\uff1a\u7edf\u4e00\u5bf9\u5e94\u70b9\u4f30\u8ba1\u3001\u76f8\u673a\u59ff\u6001\u4f30\u8ba1\u548c\u795e\u7ecf\u8f90\u5c04\u573a\u91cd\u5efa\uff0c\u5b9e\u73b0\u7aef\u5230\u7aef\u53cc\u89c6\u56fe\u65b0\u89c6\u89d2\u5408\u6210 \uff08CVPR 2024 Highlight\u8bba\u6587\uff09<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"768\" height=\"204\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-1-768x204-1.png\" alt=\"Unifying Correspondence, Pose and NeRF for Generalized Pose-free Novel View Synthesis\" class=\"wp-image-1073862\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-1-768x204-1.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-1-768x204-1-300x80.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-1-768x204-1-240x64.png 240w\" sizes=\"(max-width: 768px) 100vw, 768px\" \/><\/figure>\n\n\n\n<p>\u8bba\u6587\u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2312.07246\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/arxiv.org\/abs\/2312.07246<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>\u5728\u4f7f\u7528\u76f8\u673a\u62cd\u6444\u7684\u591a\u5f20\u4e8c\u7ef4\u56fe\u50cf\u8fdb\u884c\u4e09\u7ef4\u795e\u7ecf\u8f90\u5c04\u573a\uff08NeRF\uff09\u5efa\u6a21\u548c\u65b0\u89c6\u89d2\u6e32\u67d3\u7684\u8fc7\u7a0b\u4e2d\uff0c\u4f20\u7edf\u7684\u5904\u7406\u6d41\u7a0b\u5f80\u5f80\u5c06\u5176\u7ec6\u5206\u4e3a\u4e09\u4e2a\u5b50\u4efb\u52a1\uff1a\u56fe\u50cf\u7279\u5f81\u70b9\u5339\u914d\u3001\u76f8\u673a\u59ff\u6001\u4f30\u8ba1\u4e0e\u795e\u7ecf\u8f90\u5c04\u573a\u5efa\u6a21\uff0c\u5e76\u4e3a\u6bcf\u4e2a\u5b50\u4efb\u52a1\u8bbe\u8ba1\u4e0d\u540c\u7684\u7b97\u6cd5\u6a21\u5757\u4ee5\u5b9e\u73b0\u8be5\u5b50\u4efb\u52a1\u7684\u76ee\u6807\u3002\u7136\u800c\uff0c\u7531\u4e8e\u6bcf\u4e2a\u7b97\u6cd5\u6a21\u5757\u90fd\u662f\u72ec\u7acb\u8bbe\u8ba1\u7684\uff0c\u6bcf\u4e2a\u5b50\u4efb\u52a1\u53ef\u80fd\u4f1a\u5f15\u5165\u4e0d\u540c\u7684\u504f\u5dee\uff0c\u8fd9\u5c31\u5bfc\u81f4\u6574\u4e2a\u6d41\u7a0b\u5b58\u5728\u6f5c\u5728\u7684\u4e0d\u4e00\u81f4\u98ce\u9669\u3002\u5c24\u5176\u5728\u62cd\u6444\u56fe\u7247\u6570\u91cf\u6709\u9650\u3001\u76f8\u673a\u59ff\u6001\u53d8\u5316\u8f83\u5927\u7684\u60c5\u51b5\u4e0b\uff0c\u8be5\u6d41\u7a0b\u5f80\u5f80\u4f1a\u4f7f\u5f97\u6700\u7ec8\u65b0\u89c6\u89d2\u6e32\u67d3\u6548\u679c\u6b20\u4f73\u3002<\/p>\n\n\n\n<p>\u9488\u5bf9\u8fd9\u4e00\u95ee\u9898\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7684\u7814\u7a76\u5458\u4eec\u8ba4\u8bc6\u5230\uff0c\u4e09\u4e2a\u5b50\u4efb\u52a1\u4e4b\u95f4\u5171\u540c\u8ffd\u6c42\u7684\u6838\u5fc3\u76ee\u6807\u662f\u4ece\u4e8c\u7ef4\u56fe\u50cf\u6570\u636e\u4e2d\u7cbe\u786e\u89e3\u91ca\u5e76\u91cd\u5efa\u4e09\u7ef4\u51e0\u4f55\uff0c\u56e0\u6b64\uff0c\u5b83\u4eec\u95f4\u7684\u7edf\u4e00\u6027\u548c\u534f\u540c\u6027\u4e9f\u5f85\u52a0\u5f3a\u3002\u5728\u4ec5\u6709\u4e24\u5e45\u56fe\u50cf\u4f5c\u4e3a\u8f93\u5165\u7684\u60c5\u51b5\u4e0b\uff0c\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u4e00\u4e2a\u540d\u4e3a CoPoNeRF \u7684\u521b\u65b0\u6846\u67b6\u3002\u8be5\u6846\u67b6\u65e0\u7f1d\u6574\u5408\u4e86\u4e8c\u7ef4\u56fe\u50cf\u5bf9\u5e94\u70b9\u5339\u914d\u3001\u76f8\u673a\u76f8\u5bf9\u59ff\u6001\u4f30\u8ba1\u4e0e\u795e\u7ecf\u8f90\u5c04\u573a\u6e32\u67d3\uff0c\u5e76\u5229\u7528\u8fd9\u4e2a\u5177\u6709\u7edf\u4e00\u6027\u7684\u6846\u67b6\uff0c\u589e\u5f3a\u4e86\u5404\u5b50\u4efb\u52a1\u95f4\u7684\u534f\u540c\u6027\u3002CoPoNeRF \u6846\u67b6\u7684\u8bbe\u8ba1\u5f3a\u8c03\u7684\u662f\u4e09\u4e2a\u4efb\u52a1\u5171\u4eab\u4e00\u4e2a\u7edf\u4e00\u7684\u8868\u5f81\uff0c\u4e14\u6574\u4e2a\u6846\u67b6\u91c7\u7528\u7aef\u5230\u7aef\u7684\u65b9\u5f0f\u8fdb\u884c\u8bad\u7ec3\uff0c\u65e8\u5728\u63d0\u5347\u5b50\u4efb\u52a1\u95f4\u7684\u534f\u540c\u6027\u548c\u6574\u4f53\u6a21\u578b\u7684\u7cbe\u5ea6\u3002<\/p>\n\n\n\n<p>\u7814\u7a76\u5458\u4eec\u5728\u771f\u5b9e\u4e16\u754c\u7684\u4e24\u4e2a\u591a\u6837\u5316\u5ba4\u5185\u548c\u5ba4\u5916\u573a\u666f\u6570\u636e\u96c6\u4e0a\uff0c\u5bf9 CoPoNeRF \u6846\u67b6\u8fdb\u884c\u4e86\u5e7f\u6cdb\u7684\u8bc4\u4f30\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u8be5\u65b9\u6cd5\u76f8\u8f83\u4e8e\u5148\u524d\u7684\u65b9\u6cd5\u53d6\u5f97\u4e86\u660e\u663e\u6539\u8fdb\uff0c\u7279\u522b\u662f\u5728\u4e24\u5e45\u56fe\u50cf\u89c6\u70b9\u53d8\u5316\u8f83\u5927\u3001\u76f8\u673a\u59ff\u6001\u96be\u4ee5\u7cbe\u786e\u83b7\u5f97\u7684\u60c5\u51b5\u4e0b\u65b0\u89c6\u89d2\u6e32\u67d3\u6548\u679c\u83b7\u5f97\u6781\u5927\u63d0\u9ad8\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"393\" height=\"458\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-2.png\" alt=\"Unifying Correspondence, Pose and NeRF for Generalized Pose-free Novel View Synthesis | image collage\" class=\"wp-image-1073865\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-2.png 393w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-2-257x300.png 257w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-2-154x180.png 154w\" sizes=\"(max-width: 393px) 100vw, 393px\" \/><figcaption class=\"wp-element-caption\">\u56fe1\uff1a\u7ed9\u5b9a\u4e24\u5e45\u53ef\u80fd\u5b58\u5728\u6781\u7aef\u89c6\u70b9\u53d8\u5316\u7684\u56fe\u50cf\uff0cCoPoNeRF \u6846\u67b6\u53ef\u4ee5\u7aef\u5230\u7aef\u5730\u534f\u540c\u6267\u884c\u4e09\u4e2a\u4efb\u52a1\uff082D \u5bf9\u5e94\u70b9\u4f30\u8ba1\u3001\u76f8\u673a\u59ff\u6001\u4f30\u8ba1\u548c\u795e\u7ecf\u8f90\u5c04\u573a\u6e32\u67d3\uff09\uff0c\u4ece\u800c\u5b9e\u73b0\u9ad8\u8d28\u91cf\u7684\u65b0\u89c6\u89d2\u56fe\u50cf\u5408\u6210\u3002<\/figcaption><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"dcvc-fm-\u57fa\u4e8e\u7279\u5f81\u8c03\u5236\u7684\u89c6\u9891\u7f16\u89e3\u7801\u5668\">DCVC-FM\uff1a\u57fa\u4e8e\u7279\u5f81\u8c03\u5236\u7684\u89c6\u9891\u7f16\u89e3\u7801\u5668<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"768\" height=\"141\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-3-768x141-1.png\" alt=\"Neural Video Compression with Feature Modulation\" class=\"wp-image-1073868\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-3-768x141-1.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-3-768x141-1-300x55.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-3-768x141-1-240x44.png 240w\" sizes=\"(max-width: 768px) 100vw, 768px\" \/><\/figure>\n\n\n\n<p>\u8bba\u6587\u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2402.17414\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/arxiv.org\/abs\/2402.17414<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>\u9879\u76ee\u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/DCVC\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/github.com\/microsoft\/DCVC<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>\u76f8\u8f83\u4e8e\u5e38\u7528\u7684\u57fa\u4e8e\u6b8b\u5dee\u7f16\u7801\u7684\u89c6\u9891\u7f16\u89e3\u7801\u5668\uff0c\u57fa\u4e8e\u6761\u4ef6\u7f16\u7801\u7684\u6a21\u578b\u5c55\u73b0\u51fa\u4e86\u5de8\u5927\u7684\u6f5c\u529b\uff0c\u5e76\u5728\u8fd9\u51e0\u5e74\u53d6\u5f97\u4e86\u663e\u8457\u7684\u53d1\u5c55\u3002\u4f46\u5728\u5176\u5411\u5b9e\u7528\u5316\u8fc8\u8fdb\u7684\u8fc7\u7a0b\u4e2d\uff0c\u4ecd\u5b58\u5728\u4e24\u4e2a\u4e9f\u5f85\u89e3\u51b3\u7684\u5173\u952e\u95ee\u9898\uff1a1. \u8003\u8651\u5230\u4e0d\u540c\u5e26\u5bbd\u548c\u5b58\u50a8\u7684\u9700\u6c42\uff0c\u5982\u4f55\u8bbe\u8ba1\u4e00\u4e2a\u80fd\u591f\u652f\u6301\u53ef\u53d8\u7801\u7387\u7684\u5355\u4e00\u6a21\u578b\uff0c\u4e14\u8be5\u7801\u7387\u8303\u56f4\u9700\u8981\u8db3\u591f\u5927\uff1b2. \u89c6\u9891\u7f16\u89e3\u7801\u5668\u901a\u8fc7\u6316\u6398\u65f6\u57df\u76f8\u5173\u6027\u4ee5\u51cf\u5c0f\u89c6\u9891\u5197\u4f59\uff0c\u4f46\u8fd9\u5f80\u5f80\u5bb9\u6613\u5bfc\u81f4\u65f6\u57df\u4e0a\u7684\u8bef\u5dee\u4f20\u64ad\u3002\u56e0\u6b64\uff0c\u5982\u4f55\u907f\u514d\u8be5\u8bef\u5dee\u4f20\u64ad\u9020\u6210\u7684\u91cd\u5efa\u8d28\u91cf\u8870\u51cf\u3002\u9488\u5bf9\u8fd9\u4e24\u4e2a\u95ee\u9898\uff0c\u7814\u7a76\u5458\u4eec\u901a\u8fc7\u8bbe\u8ba1\u7279\u5f81\u8c03\u5236\u6280\u672f\uff0c\u63d0\u51fa\u4e86\u76f8\u5e94\u7684\u89e3\u51b3\u65b9\u6848\u3002<\/p>\n\n\n\n<p>\u4e3a\u4e86\u5b9e\u73b0\u5927\u8303\u56f4\u7684\u53ef\u53d8\u7801\u7387\uff0c\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4f7f\u7528\u53ef\u5b66\u4e60\u7684\u91cf\u5316\u7f29\u653e\u5668\u6765\u8c03\u6574\u6bcf\u5e27\u7684\u9690\u7279\u5f81\uff08\u5982\u56fe2\u53f3\u56fe\u6240\u793a\uff09\u3002\u5728\u8bad\u7ec3\u7684\u8fc7\u7a0b\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u8bbe\u8ba1\u4e86\u4e00\u79cd\u5747\u5300\u7684\u91cf\u5316\u53c2\u6570\u91c7\u6837\u673a\u5236\uff0c\u4ee5\u4f7f\u5f97\u7f16\u89e3\u7801\u5668\u5e94\u5bf9\u4e0d\u4e00\u6837\u7684\u7387\u5931\u771f\u6743\u8861\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"2048\" height=\"712\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1.jpg\" alt=\"Neural Video Compression with Feature Modulation | diagram\" class=\"wp-image-1073871\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1.jpg 2048w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1-300x104.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1-1024x356.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1-768x267.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1-1536x534.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-4-2048x712-1-240x83.jpg 240w\" sizes=\"(max-width: 2048px) 100vw, 2048px\" \/><figcaption class=\"wp-element-caption\">\u56fe2\uff1aDCVC-FM \u7684\u4e3b\u6846\u67b6\u56fe\uff08\u5de6\uff09\u548c\u5e27\u7f16\u7801\u6846\u67b6\u56fe\uff08\u53f3\uff09<\/figcaption><\/figure>\n\n\n\n<p>\u8be5\u8bad\u7ec3\u673a\u5236\u53ef\u4ee5\u5e2e\u52a9\u5b66\u4e60\u7ec6\u7c92\u5ea6\uff0c\u5e76\u4f18\u5316\u53ef\u8c03\u8282\u7684\u91cf\u5316\u7f29\u653e\u5668\uff0c\u4ece\u800c\u80fd\u5728\u4e00\u4e2a\u5f88\u5927\u7684\u7801\u7387\u8303\u56f4\u5185\uff0c\u6839\u636e\u7528\u6237\u9700\u6c42\u8c03\u8282\u89c6\u9891\u7f16\u7801\u8d28\u91cf\u2014\u2014\u53ef\u8c03\u7684\u89c6\u9891\u8d28\u91cf\u8303\u56f4\u5c06\u4ece\u4e4b\u524d\u76843.8 dB\u589e\u957f\u523011.4 dB\u3002\u53e6\u5916\uff0cDCVC-FM \u80fd\u591f\u52a8\u6001\u8c03\u8282\u89c6\u9891\u4e2d\u7684\u6bcf\u5e27\u8d28\u91cf\uff0c\u8be5\u529f\u80fd\u4e5f\u4f7f\u5f97\u6a21\u578b\u53ef\u4ee5\u5e94\u5bf9\u590d\u6742\u6613\u53d8\u7684\u7f51\u7edc\u5e26\u5bbd\u73af\u5883\uff0c\u56fe3\u4fbf\u5c55\u793a\u4e86 DCVC-FM \u5728\u5e94\u5bf9\u53d8\u5316\u5e26\u5bbd\u65b9\u9762\u7684\u80fd\u529b\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"670\" height=\"368\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-5.png\" alt=\"Neural Video Compression with Feature Modulation | line chart\" class=\"wp-image-1073874\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-5.png 670w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-5-300x165.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-5-240x132.png 240w\" sizes=\"(max-width: 670px) 100vw, 670px\" \/><figcaption class=\"wp-element-caption\">\u56fe3\uff1a\u7801\u7387\u63a7\u5236\u793a\u610f<\/figcaption><\/figure>\n\n\n\n<p>\u9488\u5bf9\u65f6\u57df\u8bef\u5dee\u4f20\u64ad\u95ee\u9898\uff0c\u5df2\u6709\u7684\u6a21\u578b\u666e\u904d\u91c7\u7528\u9891\u7e41\u63d2\u5165\u9ad8\u8d28\u91cf\u5e27\u5185\u7f16\u7801\u5e27\u7684\u65b9\u6cd5\u6765\u5207\u65ad\u8bef\u5dee\u4f20\u64ad\u3002\u5c3d\u7ba1\u8be5\u65b9\u6cd5\u53ef\u4ee5\u6709\u6548\u904f\u5236\u8bef\u5dee\u7684\u6269\u6563\uff0c\u4f46\u7531\u4e8e\u5e27\u5185\u7f16\u7801\u7684\u6548\u7387\u6781\u4f4e\uff0c\u6574\u4f53\u7684\u89c6\u9891\u7f16\u7801\u6548\u7387\u5c06\u663e\u8457\u964d\u4f4e\u3002\u4e3a\u6b64\uff0c\u7814\u7a76\u5458\u4eec\u8bbe\u8ba1\u4e86\u4e00\u79cd\u7279\u5f81\u66f4\u65b0\u673a\u5236\u2014\u2014\u5f53\u76f8\u90bb\u5e27\u7684\u7279\u5f81\u5728\u65f6\u57df\u4f20\u64ad\u65f6\uff0c\u7814\u7a76\u5458\u4eec\u5c06\u5468\u671f\u6027\u5730\u4f7f\u7528\u8f85\u52a9\u6027\u7684\u7279\u5f81\u63d0\u53d6\u5668\u53bb\u8c03\u6574\u5e76\u66f4\u65b0\u88ab\u4f20\u64ad\u7684\u7279\u5f81\u3002\u8fd9\u4e00\u673a\u5236\u4f7f\u5f97 DCVC-FMDCVC-FM \u5728\u5904\u7406\u975e\u5e38\u957f\u7684\u89c6\u9891\u65f6\u4ecd\u80fd\u4fdd\u6301\u6781\u9ad8\u7684\u538b\u7f29\u6548\u7387\u3002\u5982\u56fe4\u6240\u793a\uff0cDCVC-FM \u5728\u538b\u7f29\u6027\u80fd\u65b9\u9762\u5df2\u7ecf\u8d85\u8fc7\u4e86\u6b63\u5728\u7814\u53d1\u7684\u4e0b\u4e00\u4ee3\u4f20\u7edf\u89c6\u9891\u7f16\u7801\u5668\u6807\u51c6\u7684\u539f\u578b ECM\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"634\" height=\"397\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-6.png\" alt=\"Neural Video Compression with Feature Modulation | chart\" class=\"wp-image-1073877\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-6.png 634w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-6-300x188.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-6-240x150.png 240w\" sizes=\"(max-width: 634px) 100vw, 634px\" \/><figcaption class=\"wp-element-caption\">\u56fe4\uff1a\u538b\u7f29\u6027\u80fd\u5bf9\u6bd4<\/figcaption><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"mvgformer-\u7528\u4e8e3d\u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\u7684\u591a\u89c6\u89d2\u51e0\u4f55transformers\">MVGFormer\uff1a\u7528\u4e8e3D\u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\u7684\u591a\u89c6\u89d2\u51e0\u4f55Transformers<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"768\" height=\"120\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-7-768x120-1.png\" alt=\"Multiple View Geometry Transformers for 3D Human Post Estimation\" class=\"wp-image-1073880\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-7-768x120-1.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-7-768x120-1-300x47.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-7-768x120-1-240x38.png 240w\" sizes=\"(max-width: 768px) 100vw, 768px\" \/><\/figure>\n\n\n\n<p>\u8bba\u6587\u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2311.10983\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/arxiv.org\/abs\/2311.10983<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>\u9879\u76ee\u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/XunshanMan\/MVGFormer\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/github.com\/XunshanMan\/MVGFormer<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>\u5728\u8ba1\u7b97\u673a\u89c6\u89c9\u548c\u673a\u5668\u5b66\u4e60\u9886\u57df\uff0c3D \u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\u662f\u4e00\u9879\u5177\u6709\u6311\u6218\u6027\u7684\u4efb\u52a1\uff0c\u56e0\u4e3a\u5b83\u8981\u6c42\u7b97\u6cd5\u80fd\u591f\u4ece\u56fe\u50cf\u6216\u89c6\u9891\u4e2d\u51c6\u786e\u5730\u9884\u6d4b\u51fa\u4eba\u4f53\u5173\u8282\u7684 3D \u7a7a\u95f4\u4f4d\u7f6e\u3002\u8fd9\u9879\u6280\u672f\u5bf9\u4e8e\u6df1\u5165\u7406\u89e3\u4eba\u7c7b\u884c\u4e3a\u3001\u589e\u5f3a\u73b0\u5b9e\u3001\u865a\u62df\u73b0\u5b9e\u3001\u6e38\u620f\u52a8\u753b\u4ee5\u53ca\u4eba\u673a\u4ea4\u4e92\u7b49\u9886\u57df\u5177\u6709\u91cd\u8981\u7684\u5e94\u7528\u6f5c\u529b\u3002\u7136\u800c\uff0c\u73b0\u6709\u7684\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u65b9\u6cd5\u5728\u5904\u7406\u591a\u89c6\u89d2\u6570\u636e\u65f6\uff0c\u5e38\u5e38\u9762\u4e34\u906e\u6321\u95ee\u9898\u548c\u89c6\u89d2\u53d8\u5316\u6240\u5e26\u6765\u7684\u6311\u6218\uff0c\u7279\u522b\u662f\u5728\u51c6\u786e\u89e3\u6790\u51e0\u4f55\u4fe1\u606f\u65b9\u9762\u4ecd\u663e\u4e0d\u8db3\u3002<\/p>\n\n\n\n<p>\u5c3d\u7ba1\u76ee\u524d\u7684\u7814\u7a76\u5df2\u7ecf\u901a\u8fc7\u4f7f\u7528\u795e\u7ecf\u7f51\u7edc\uff0c\u5728 3D \u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\u4e0a\u53d6\u5f97\u4e86\u8fdb\u6b65\uff0c\u4f46\u8fd9\u4e9b\u65b9\u6cd5\u5728\u591a\u89c6\u89d2\u8bbe\u7f6e\u4e2d\uff0c\u5c24\u5176\u662f\u5728\u8bad\u7ec3\u9636\u6bb5\u672a\u9047\u5230\u7684\u65b0\u89c6\u89d2\u548c\u73af\u5883\u6761\u4ef6\u4e0b\uff0c\u5176\u6027\u80fd\u8868\u73b0\u5f80\u5f80\u4e0d\u5c3d\u5982\u4eba\u610f\u3002\u6b64\u5916\uff0c\u73b0\u6709\u7684\u57fa\u4e8e Transformers \u7684\u7aef\u5230\u7aef\u5b66\u4e60\u65b9\u6cd5\u867d\u7136\u80fd\u591f\u5904\u7406\u906e\u6321\u95ee\u9898\uff0c\u4f46\u5728\u65b0\u573a\u666f\u4e0b\u7684\u6cdb\u5316\u6027\u80fd\u5f80\u5f80\u8f83\u5dee\u3002\u56e0\u6b64\uff0c\u5982\u4f55\u5728\u4fdd\u6301\u7aef\u5230\u7aef\u5b66\u4e60\u4f18\u52bf\u7684\u57fa\u7840\u4e0a\uff0c\u63d0\u5347\u6a21\u578b\u5bf9\u4e8e\u65b0\u89c6\u89d2\u548c\u906e\u6321\u60c5\u51b5\u7684\u6cdb\u5316\u80fd\u529b\uff0c\u662f\u5f53\u524d\u7814\u7a76\u9762\u4e34\u7684\u5173\u952e\u8bfe\u9898\u3002<\/p>\n\n\n\n<p>\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u4e00\u79cd\u65b0\u7684\u65b9\u6cd5\uff0c\u5176\u4e3b\u8981\u601d\u60f3\u53ef\u4ee5\u6982\u62ec\u4e3a\u4ee5\u4e0b\u4e09\u4e2a\u90e8\u5206\uff1a<\/p>\n\n\n\n<p>1\uff09\u6df7\u5408\u6a21\u578b\u8bbe\u8ba1\uff1a\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u4e00\u79cd\u65b0\u9896\u7684\u6df7\u5408\u6a21\u578b MVGFormer\uff0c\u8be5\u6a21\u578b\u7ed3\u5408\u4e86\u51e0\u4f55\u6a21\u5757\u548c\u5916\u89c2\u6a21\u5757\u3002\u51e0\u4f55\u6a21\u5757\u8d1f\u8d23\u5904\u7406\u6240\u6709\u4e0e\u89c6\u89d2\u76f8\u5173\u7684 3D \u4efb\u52a1\uff0c\u800c\u5916\u89c2\u6a21\u5757\u5219\u4e13\u6ce8\u4e8e\u4ece\u56fe\u50cf\u4fe1\u53f7\u4e2d\u4f30\u8ba1 2D \u59ff\u6001\u3002<\/p>\n\n\n\n<p>2\uff09\u8fed\u4ee3\u7ec6\u5316\u8fc7\u7a0b\uff1aMVGFormer \u901a\u8fc7\u8fed\u4ee3\u7684\u65b9\u5f0f\uff0c\u4ea4\u66ff\u4f7f\u7528\u5916\u89c2\u6a21\u5757\u548c\u51e0\u4f55\u6a21\u5757\u6765\u7ec6\u5316 2D \u548c 3D \u59ff\u6001\u3002\u5916\u89c2\u6a21\u5757\u57fa\u4e8e\u56fe\u50cf\u7279\u5f81\u9884\u6d4b2D\u59ff\u6001\uff0c\u51e0\u4f55\u6a21\u5757\u5229\u7528\u4e09\u89d2\u6d4b\u91cf\u6cd5\u4ece 2D \u59ff\u6001\u4e2d\u6062\u590d 3D \u59ff\u6001\u3002<\/p>\n\n\n\n<p>3\uff09\u7aef\u5230\u7aef\u8bad\u7ec3\u4e0e\u8bc4\u4f30\uff1aMVGFormer \u652f\u6301\u7aef\u5230\u7aef\u7684\u8bad\u7ec3\uff0c\u56e0\u6b64\u7814\u7a76\u5458\u4eec\u80fd\u901a\u8fc7\u4e00\u7cfb\u5217\u8bad\u7ec3\u76ee\u6807\u548c\u635f\u5931\u51fd\u6570\u6765\u4f18\u5316\u6a21\u578b\u3002\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u6a21\u578b\u5c06\u5b66\u4e60\u5982\u4f55\u51c6\u786e\u5730\u4ece\u4e00\u7ec4\u521d\u59cb\u5316\u67e5\u8be2\u4e2d\u91cd\u5efa 3D \u4eba\u4f53\u59ff\u6001\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"339\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-8-1024x339.png\" alt=\"Multiple View Geometry Transformers for 3D Human Post Estimation | diagram\" class=\"wp-image-1073883\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-8-1024x339.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-8-300x99.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-8-768x254.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-8-240x79.png 240w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-8.png 1169w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\">\u56fe5\uff1aMVGFormer \u6a21\u578b\u7ed3\u6784<\/figcaption><\/figure>\n\n\n\n<p>\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u76f8\u8f83\u4e8e\u76ee\u524d\u9886\u57df\u5185\u7684\u524d\u6cbf\u65b9\u6cd5\uff0cMVGFormer \u5728\u591a\u79cd\u6d4b\u8bd5\u6761\u4ef6\u4e0b\u5747\u5c55\u73b0\u51fa\u4e86\u5353\u8d8a\u7684\u6027\u80fd\uff0c\u7279\u522b\u662f\u5728\u90a3\u4e9b\u8bad\u7ec3\u9636\u6bb5\u672a\u66fe\u9047\u5230\u7684\u65b0\u89c6\u89d2\u4e0b\uff0c\u5176\u8868\u73b0\u5c24\u4e3a\u7a81\u51fa\u3002\u8fd9\u4e00\u6210\u5c31\u4e3b\u8981\u5f97\u76ca\u4e8e MVGFormer \u5f3a\u5927\u7684\u6cdb\u5316\u80fd\u529b\uff0c\u5373\u4fbf\u5728\u5168\u65b0\u7684\u89c6\u89d2\u6761\u4ef6\u4e0b\uff0c\u5b83\u4ecd\u80fd\u4fdd\u6301\u5bf9 3D \u4eba\u4f53\u59ff\u6001\u7684\u9ad8\u7cbe\u5ea6\u4f30\u8ba1\uff0c\u5bf9\u4e8e\u9700\u8981\u5728\u591a\u79cd\u89c6\u89d2\u4e0b\u7406\u89e3\u548c\u9884\u6d4b\u4eba\u4f53\u52a8\u4f5c\u7684\u5e94\u7528\u573a\u666f\u5177\u6709\u91cd\u8981\u610f\u4e49\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1105\" height=\"656\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-9.png\" alt=\"Multiple View Geometry Transformers for 3D Human Post Estimation | diagram\" class=\"wp-image-1073886\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-9.png 1105w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-9-300x178.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-9-1024x608.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-9-768x456.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-9-240x142.png 240w\" sizes=\"(max-width: 1105px) 100vw, 1105px\" \/><figcaption class=\"wp-element-caption\">\u56fe6\uff1aMVGFormer \u5c06\u521d\u59cb\u5316\u67e5\u8be2\u6062\u590d\u51fa 3D \u4eba\u4f53\u59ff\u6001\u7684\u8fc7\u7a0b<\/figcaption><\/figure>\n\n\n\n<p>\u6b64\u5916\uff0cMVGFormer \u7684\u6846\u67b6\u8fd8\u5177\u6709\u901a\u7528\u6027\uff0c\u53ef\u4ee5\u5e94\u7528\u4e8e\u5176\u4ed6\u5173\u952e\u70b9\u4f30\u8ba1\u4efb\u52a1\uff0c\u5982\u5f62\u72b6\u3001\u624b\u90e8\u548c\u9762\u90e8\u4f30\u8ba1\u3002\u672a\u6765\uff0c\u7814\u7a76\u5458\u4eec\u4e5f\u53ef\u80fd\u4f1a\u63a2\u7d22\u5c06\u8be5\u67b6\u6784\u6269\u5c55\u5230\u57fa\u4e8e\u89c6\u9891\u7684\u7cfb\u7edf\u4e2d\uff0c\u4ee5\u5229\u7528\u65f6\u95f4\u4fe1\u606f\u5b9e\u73b0\u66f4\u7a33\u5065\u7684\u8ffd\u8e2a\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"\u6587\u672c\u5206\u7ec4\u9002\u914d\u5668-\u5c06\u6587\u672c\u5e03\u5c40\u5206\u6790\u80fd\u529b\u88c5\u914d\u5728\u4efb\u610f\u6587\u672c\u68c0\u6d4b\u5668\u4e0a\">\u6587\u672c\u5206\u7ec4\u9002\u914d\u5668\uff1a\u5c06\u6587\u672c\u5e03\u5c40\u5206\u6790\u80fd\u529b\u88c5\u914d\u5728\u4efb\u610f\u6587\u672c\u68c0\u6d4b\u5668\u4e0a<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"768\" height=\"147\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-10-768x147-1.png\" alt=\"Text Grouping Adapter: Adapting Pre-trained Text Detector for Layout Analysis\" class=\"wp-image-1073889\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-10-768x147-1.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-10-768x147-1-300x57.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-10-768x147-1-240x46.png 240w\" sizes=\"(max-width: 768px) 100vw, 768px\" \/><\/figure>\n\n\n\n<p>\u8bba\u6587\u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2405.07481\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/arxiv.org\/abs\/2405.07481<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>\u968f\u7740\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u53d1\u5c55\uff0c\u573a\u666f\u6587\u672c\u68c0\u6d4b\u548c\u8bc6\u522b\u9886\u57df\u5df2\u53d6\u5f97\u4e86\u91cd\u5927\u8fdb\u6b65\u3002\u7136\u800c\uff0c\u5982\u4f55\u5c06\u68c0\u6d4b\u5230\u7684\u4f17\u591a\u6587\u672c\u533a\u57df\u8fdb\u884c\u7b26\u5408\u8bed\u4e49\u7684\u5206\u7ec4\u5e76\u6784\u5efa\u6bb5\u843d\u7ed3\u6784\u2014\u2014\u573a\u666f\u6587\u672c\u5e03\u5c40\u5206\u6790\u95ee\u9898\uff0c\u4ecd\u9762\u4e34\u7740\u8bf8\u591a\u6311\u6218\u3002\u8fd9\u4e9b\u95ee\u9898\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\u6807\u6ce8\u6570\u636e\u96c6\u7f3a\u4e4f\u3001\u8f83\u96be\u5229\u7528\u5df2\u6709\u7684\u6587\u672c\u68c0\u6d4b\u5668\u3001\u7f3a\u5c11\u7aef\u5230\u7aef\u6a21\u578b\u7b49\u3002<\/p>\n\n\n\n<p>\u7814\u7a76\u5458\u4eec\u53d1\u73b0\uff0c\u573a\u666f\u6587\u672c\u5e03\u5c40\u5206\u6790\u7684\u7ed3\u679c\u6781\u5927\u5730\u4f9d\u8d56\u4e8e\u6587\u672c\u68c0\u6d4b\u7684\u51c6\u786e\u6027\uff0c\u4f46\u73b0\u6709\u7684\u6587\u672c\u5e03\u5c40\u5206\u6790\u65b9\u6cd5\u8981\u6c42\u4ece\u5934\u5f00\u59cb\u8054\u5408\u6587\u672c\u5e03\u5c40\u5206\u6790\uff0c\u4e0e\u6587\u672c\u68c0\u6d4b\u5668\u8fdb\u884c\u8054\u5408\u8bad\u7ec3\uff0c\u8fd9\u79cd\u65b9\u6cd5\u5f80\u5f80\u5bfc\u81f4\u6587\u672c\u68c0\u6d4b\u5668\u8bad\u7ec3\u4e0d\u5145\u5206\u3002<\/p>\n\n\n\n<p>\u57fa\u4e8e\u8fd9\u4e00\u53d1\u73b0\uff0c\u7814\u7a76\u5458\u4eec\u5229\u7528\u5728\u5927\u91cf\u6587\u672c\u68c0\u6d4b\u6570\u636e\u96c6\u4e0a\u8bad\u7ec3\u5f97\u5230\u7684\u9884\u8bad\u7ec3\u6587\u672c\u68c0\u6d4b\u5668\uff0c\u6210\u529f\u89e3\u51b3\u4e86\u56e0\u7f51\u7edc\u7ed3\u6784\u5dee\u5f02\u548c\u8f93\u51fa\u683c\u5f0f\u4e0d\u540c\u6240\u5e26\u6765\u7684\u6311\u6218\uff0c\u5e76\u4e14\u8ba9\u6a21\u578b\u5728\u51bb\u7ed3\u6240\u6709\u539f\u6709\u6743\u91cd\u7684\u60c5\u51b5\u4e0b\u80fd\u591f\u6b63\u5e38\u8fdb\u884c\u6587\u672c\u5e03\u5c40\u5206\u6790\u3002\u6700\u7ec8\uff0c\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u901a\u7528\u7684\u6587\u672c\u5206\u7ec4\u9002\u914d\u5668\uff08Text Grouping Adapter\uff0cTGA\uff09\uff0c\u663e\u8457\u63d0\u5347\u4e86\u6587\u672c\u5e03\u5c40\u5206\u6790\u6027\u80fd\uff0c\u5e76\u5927\u5e45\u52a0\u901f\u4e86\u8bad\u7ec3\u6d41\u7a0b\u3002<\/p>\n\n\n\n<p>\u56fe7\u5de6\u56fe\u5c55\u793a\u4e86\u88c5\u914d\u4e86 TGA \u7684\u6a21\u578b\u6574\u4f53\u6846\u67b6\uff0c\u5176\u4e2d\u539f\u6587\u672c\u68c0\u6d4b\u5668\u6a21\u5757\u53ef\u9009\u62e9\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u51bb\u7ed3\u6240\u6709\u6743\u91cd\u3002\u53f3\u56fe\u5219\u5c55\u793a\u4e86 TGA \u7684\u5185\u90e8\u7ed3\u6784\uff1a\u9996\u5148\uff0c\u6587\u672c\u533a\u57df\u5c06\u88ab\u8f6c\u5316\u6210\u5b9e\u4f8b\u63a9\u7801\uff0c\u4ece\u800c\u5f97\u5230\u6587\u672c\u5b9e\u4f8b\u7279\u5f81\uff1b\u63a5\u7740\uff0c\u9884\u6d4b\u6587\u672c\u7ec4\u5408\u63a9\u7801\uff0c\u786e\u4fdd\u5c5e\u4e8e\u540c\u4e00\u7ec4\u7684\u6587\u672c\u5b9e\u4f8b\u5177\u6709\u76f8\u4f3c\u7684\u7279\u5f81\uff1b\u6700\u7ec8\uff0c\u8f93\u51fa\u6587\u672c\u5b9e\u4f8b\u7684\u5173\u7cfb\u77e9\u9635\uff0c\u4ee5\u6b64\u5f62\u6210\u6587\u672c\u7ec4\u5408\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1320\" height=\"541\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-11.png\" alt=\"Text Grouping Adapter: Adapting Pre-trained Text Detector for Layout Analysis | diagram\" class=\"wp-image-1073892\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-11.png 1320w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-11-300x123.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-11-1024x420.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-11-768x315.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2024\/08\/new-arrival-in-research-12-11-240x98.png 240w\" sizes=\"(max-width: 1320px) 100vw, 1320px\" \/><figcaption class=\"wp-element-caption\">\u56fe7\uff1a\u88c5\u914d\u6709 TGA \u7684\u6a21\u578b\u6574\u4f53\u6846\u67b6\uff08\u5de6\uff09\u548c TGA \u5185\u90e8\u7ed3\u6784\uff08\u53f3\uff09<\/figcaption><\/figure>\n\n\n\n<p>\u5177\u4f53\u800c\u8a00\uff0cTGA \u5305\u62ec\u4e24\u4e2a\u91cd\u8981\u6a21\u5757\uff1a\u6587\u672c\u533a\u57df\u7279\u5f81\u7ec4\u88c5\u6a21\u5757\uff08Text Instance Feature Assembling\uff09\u548c\u6587\u672c\u7ec4\u5408\u63a9\u7801\u9884\u6d4b\u6a21\u5757\uff08Group Mask Prediction\uff09\u3002\u6587\u672c\u533a\u57df\u7279\u5f81\u7ec4\u88c5\u6a21\u5757\u4e3b\u8981\u628a\u4e0d\u540c\u6587\u672c\u68c0\u6d4b\u5668\u7684\u8f93\u51fa\u7edf\u4e00\u6210\u4e3a\u6587\u672c\u533a\u57df\u63a9\u7801\u7684\u5f62\u5f0f\uff0c\u5e76\u4e14\u901a\u8fc7\u63a9\u7801\u548c\u56fe\u7247\u50cf\u7d20\u5d4c\u5165\u6765\u7ec4\u88c5\u6210\u6587\u672c\u533a\u57df\u7279\u5f81\uff0c\u8fd9\u4e9b\u7279\u5f81\u968f\u540e\u4f1a\u88ab\u9001\u5165\u6587\u672c\u7ec4\u5408\u63a9\u7801\u9884\u6d4b\u6a21\u5757\u4ee5\u53ca\u6700\u7ec8\u7684\u6587\u672c\u5173\u7cfb\u9884\u6d4b\u6a21\u5757\u3002\u800c\u6587\u672c\u7ec4\u5408\u63a9\u7801\u9884\u6d4b\u6a21\u5757\u5219\u901a\u8fc7\u9884\u6d4b\u6587\u672c\u7ec4\u5408\u7684\u5171\u540c\u63a9\u7801\uff0c\u786e\u4fdd\u5c5e\u4e8e\u540c\u4e00\u6587\u672c\u7ec4\u5408\u7684\u6587\u672c\u533a\u57df\u5177\u6709\u76f8\u4f3c\u7684\u7279\u5f81\uff0c\u4ece\u800c\u5728\u6587\u672c\u5b9e\u4f8b\u7684\u5173\u7cfb\u77e9\u9635\u4e2d\u9884\u6d4b\u51fa\u6b63\u786e\u7684\u5173\u7cfb\u3002<\/p>\n\n\n\n<p>\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0cTGA \u80fd\u4e0e\u591a\u79cd\u4e0d\u540c\u7f51\u7edc\u7ed3\u6784\u4e0e\u8f93\u51fa\u683c\u5f0f\u7684\u6587\u672c\u68c0\u6d4b\u5668\u517c\u5bb9\u3002\u6b64\u5916\uff0cTGA \u8fd8\u80fd\u591f\u4e0e\u7aef\u5230\u7aef\u6587\u672c\u8bc6\u522b\u6a21\u578b\uff08Text Spotting Model\uff09\u7ed3\u5408\u4f7f\u7528\uff0c\u5b9e\u73b0\u6587\u672c\u68c0\u6d4b\u3001\u6587\u672c\u8bc6\u522b\u4ee5\u53ca\u6587\u672c\u5e03\u5c40\u5206\u6790\u7ed3\u679c\u7684\u540c\u6b65\u8f93\u51fa\u3002\u901a\u8fc7\u51bb\u7ed3\u6587\u672c\u68c0\u6d4b\u5668\uff0c\u6a21\u578b\u80fd\u5728\u4ec5\u6709\u4e4b\u524d\u5de5\u4f5c13%\u7684\u53ef\u8bad\u7ec3\u53c2\u6570\u7684\u60c5\u51b5\u4e0b\uff0c\u4ecd\u80fd\u591f\u5f97\u5230\u9886\u5148\u7684\u6587\u672c\u68c0\u6d4b\u548c\u6587\u672c\u5e03\u5c40\u5206\u6790\u7ed3\u679c\u3002\u7814\u7a76\u5458\u4eec\u9488\u5bf9 TGA \u635f\u5931\u51fd\u6570\u7684\u4e00\u7cfb\u5217\u6d88\u878d\u5b9e\u9a8c\u8fdb\u4e00\u6b65\u8868\u660e\uff0cTGA \u6709\u6548\u89e3\u51b3\u4e86\u6b63\u8d1f\u6837\u672c\u4e0d\u5e73\u8861\u7684\u95ee\u9898\u3002<\/p>\n\n\n\n<p>\u6587\u672c\u5e03\u5c40\u5206\u6790\u80cc\u540e\u7684\u6838\u5fc3\u95ee\u9898\u5728\u4e8e\uff0c\u89c6\u89c9\u6a21\u578b\u5982\u4f55\u51c6\u786e\u5224\u65ad\u89c6\u89c9\u5bf9\u8c61\u95f4\u7684\u5173\u7cfb\u4ee5\u53ca\u5b83\u4eec\u5728\u573a\u666f\u4e2d\u7684\u62d3\u6251\u7ed3\u6784\u3002\u672a\u6765\uff0c\u7814\u7a76\u5458\u4eec\u5c06\u6301\u7eed\u5173\u6ce8\u8fd9\u4e00\u95ee\u9898\uff0c\u5e76\u5728 Windows UI \u573a\u666f\u548c\u81ea\u7136\u56fe\u50cf\u5206\u6790\u65b9\u9762\u8fdb\u884c\u66f4\u52a0\u6df1\u5165\u7684\u63a2\u7d22\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u7f16\u8005\u6309\uff1a\u6b22\u8fce\u9605\u8bfb\u201c\u79d1\u7814\u4e0a\u65b0\u201d\u680f\u76ee\uff01\u201c\u79d1\u7814\u4e0a\u65b0\u201d\u6c47\u805a\u4e86\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u6700\u65b0\u7684\u521b\u65b0\u6210\u679c\u4e0e\u79d1\u7814\u52a8\u6001\u3002\u5728\u8fd9\u91cc\uff0c\u4f60\u53ef\u4ee5\u5feb\u901f\u6d4f\u89c8\u7814\u7a76\u9662\u7684\u4eae\u70b9\u8d44\u8baf\uff0c\u4fdd\u6301\u5bf9\u524d\u6cbf\u9886\u57df\u7684\u654f\u9510\u55c5\u89c9\uff0c\u540c\u65f6\u4e5f\u80fd\u627e\u5230\u5148\u8fdb\u5b9e\u7528\u7684\u5f00\u6e90\u5de5\u5177\u3002 \u672c\u5468\uff0c\u4eba\u5de5\u667a\u80fd\u9886\u57df\u6700\u5177\u5b66\u672f\u5f71\u54cd\u529b\u7684\u9876\u7ea7\u4f1a\u8bae\u4e4b\u4e00\uff0cCVPR \u5927\u4f1a\u5728\u7f8e\u56fd\u897f\u96c5\u56fe\u4e3e\u529e\u3002\u56e0\u6b64\uff0c\u8fd9\u671f\u7684\u201c\u79d1\u7814\u4e0a\u65b0\u201d\u4e3a\u5927\u5bb6\u5e26\u6765\u4e86\u56db\u7bc7\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5165\u9009 CVPR 2024 \u7684\u7cbe\u9009\u8bba\u6587\u89e3\u8bfb\uff0c\u6d89\u53ca\u9886\u57df\u6db5\u76d6\u65b0\u89c6\u89d2\u56fe\u50cf\u5408\u6210\u30013D \u4eba\u4f53\u59ff\u6001\u4f30\u8ba1\u3001\u89c6\u9891\u7f16\u89e3\u7801\u5668\u3001\u6587\u672c\u5e03\u5c40\u5206\u6790\u7b49\u591a\u4e2a\u76f8\u5173\u4e3b\u9898\u3002 \u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2312.07246 (opens in new tab) \u5728\u4f7f\u7528\u76f8\u673a\u62cd\u6444\u7684\u591a\u5f20\u4e8c\u7ef4\u56fe\u50cf\u8fdb\u884c\u4e09\u7ef4\u795e\u7ecf\u8f90\u5c04\u573a\uff08NeRF\uff09\u5efa\u6a21\u548c\u65b0\u89c6\u89d2\u6e32\u67d3\u7684\u8fc7\u7a0b\u4e2d\uff0c\u4f20\u7edf\u7684\u5904\u7406\u6d41\u7a0b\u5f80\u5f80\u5c06\u5176\u7ec6\u5206\u4e3a\u4e09\u4e2a\u5b50\u4efb\u52a1\uff1a\u56fe\u50cf\u7279\u5f81\u70b9\u5339\u914d\u3001\u76f8\u673a\u59ff\u6001\u4f30\u8ba1\u4e0e\u795e\u7ecf\u8f90\u5c04\u573a\u5efa\u6a21\uff0c\u5e76\u4e3a\u6bcf\u4e2a\u5b50\u4efb\u52a1\u8bbe\u8ba1\u4e0d\u540c\u7684\u7b97\u6cd5\u6a21\u5757\u4ee5\u5b9e\u73b0\u8be5\u5b50\u4efb\u52a1\u7684\u76ee\u6807\u3002\u7136\u800c\uff0c\u7531\u4e8e\u6bcf\u4e2a\u7b97\u6cd5\u6a21\u5757\u90fd\u662f\u72ec\u7acb\u8bbe\u8ba1\u7684\uff0c\u6bcf\u4e2a\u5b50\u4efb\u52a1\u53ef\u80fd\u4f1a\u5f15\u5165\u4e0d\u540c\u7684\u504f\u5dee\uff0c\u8fd9\u5c31\u5bfc\u81f4\u6574\u4e2a\u6d41\u7a0b\u5b58\u5728\u6f5c\u5728\u7684\u4e0d\u4e00\u81f4\u98ce\u9669\u3002\u5c24\u5176\u5728\u62cd\u6444\u56fe\u7247\u6570\u91cf\u6709\u9650\u3001\u76f8\u673a\u59ff\u6001\u53d8\u5316\u8f83\u5927\u7684\u60c5\u51b5\u4e0b\uff0c\u8be5\u6d41\u7a0b\u5f80\u5f80\u4f1a\u4f7f\u5f97\u6700\u7ec8\u65b0\u89c6\u89d2\u6e32\u67d3\u6548\u679c\u6b20\u4f73\u3002 \u9488\u5bf9\u8fd9\u4e00\u95ee\u9898\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7684\u7814\u7a76\u5458\u4eec\u8ba4\u8bc6\u5230\uff0c\u4e09\u4e2a\u5b50\u4efb\u52a1\u4e4b\u95f4\u5171\u540c\u8ffd\u6c42\u7684\u6838\u5fc3\u76ee\u6807\u662f\u4ece\u4e8c\u7ef4\u56fe\u50cf\u6570\u636e\u4e2d\u7cbe\u786e\u89e3\u91ca\u5e76\u91cd\u5efa\u4e09\u7ef4\u51e0\u4f55\uff0c\u56e0\u6b64\uff0c\u5b83\u4eec\u95f4\u7684\u7edf\u4e00\u6027\u548c\u534f\u540c\u6027\u4e9f\u5f85\u52a0\u5f3a\u3002\u5728\u4ec5\u6709\u4e24\u5e45\u56fe\u50cf\u4f5c\u4e3a\u8f93\u5165\u7684\u60c5\u51b5\u4e0b\uff0c\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u4e00\u4e2a\u540d\u4e3a CoPoNeRF \u7684\u521b\u65b0\u6846\u67b6\u3002\u8be5\u6846\u67b6\u65e0\u7f1d\u6574\u5408\u4e86\u4e8c\u7ef4\u56fe\u50cf\u5bf9\u5e94\u70b9\u5339\u914d\u3001\u76f8\u673a\u76f8\u5bf9\u59ff\u6001\u4f30\u8ba1\u4e0e\u795e\u7ecf\u8f90\u5c04\u573a\u6e32\u67d3\uff0c\u5e76\u5229\u7528\u8fd9\u4e2a\u5177\u6709\u7edf\u4e00\u6027\u7684\u6846\u67b6\uff0c\u589e\u5f3a\u4e86\u5404\u5b50\u4efb\u52a1\u95f4\u7684\u534f\u540c\u6027\u3002CoPoNeRF \u6846\u67b6\u7684\u8bbe\u8ba1\u5f3a\u8c03\u7684\u662f\u4e09\u4e2a\u4efb\u52a1\u5171\u4eab\u4e00\u4e2a\u7edf\u4e00\u7684\u8868\u5f81\uff0c\u4e14\u6574\u4e2a\u6846\u67b6\u91c7\u7528\u7aef\u5230\u7aef\u7684\u65b9\u5f0f\u8fdb\u884c\u8bad\u7ec3\uff0c\u65e8\u5728\u63d0\u5347\u5b50\u4efb\u52a1\u95f4\u7684\u534f\u540c\u6027\u548c\u6574\u4f53\u6a21\u578b\u7684\u7cbe\u5ea6\u3002 \u7814\u7a76\u5458\u4eec\u5728\u771f\u5b9e\u4e16\u754c\u7684\u4e24\u4e2a\u591a\u6837\u5316\u5ba4\u5185\u548c\u5ba4\u5916\u573a\u666f\u6570\u636e\u96c6\u4e0a\uff0c\u5bf9 CoPoNeRF \u6846\u67b6\u8fdb\u884c\u4e86\u5e7f\u6cdb\u7684\u8bc4\u4f30\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u8be5\u65b9\u6cd5\u76f8\u8f83\u4e8e\u5148\u524d\u7684\u65b9\u6cd5\u53d6\u5f97\u4e86\u660e\u663e\u6539\u8fdb\uff0c\u7279\u522b\u662f\u5728\u4e24\u5e45\u56fe\u50cf\u89c6\u70b9\u53d8\u5316\u8f83\u5927\u3001\u76f8\u673a\u59ff\u6001\u96be\u4ee5\u7cbe\u786e\u83b7\u5f97\u7684\u60c5\u51b5\u4e0b\u65b0\u89c6\u89d2\u6e32\u67d3\u6548\u679c\u83b7\u5f97\u6781\u5927\u63d0\u9ad8\u3002 \u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2402.17414 (opens in new tab) \u9879\u76ee\u94fe\u63a5\uff1ahttps:\/\/github.com\/microsoft\/DCVC (opens in new tab) \u76f8\u8f83\u4e8e\u5e38\u7528\u7684\u57fa\u4e8e\u6b8b\u5dee\u7f16\u7801\u7684\u89c6\u9891\u7f16\u89e3\u7801\u5668\uff0c\u57fa\u4e8e\u6761\u4ef6\u7f16\u7801\u7684\u6a21\u578b\u5c55\u73b0\u51fa\u4e86\u5de8\u5927\u7684\u6f5c\u529b\uff0c\u5e76\u5728\u8fd9\u51e0\u5e74\u53d6\u5f97\u4e86\u663e\u8457\u7684\u53d1\u5c55\u3002\u4f46\u5728\u5176\u5411\u5b9e\u7528\u5316\u8fc8\u8fdb\u7684\u8fc7\u7a0b\u4e2d\uff0c\u4ecd\u5b58\u5728\u4e24\u4e2a\u4e9f\u5f85\u89e3\u51b3\u7684\u5173\u952e\u95ee\u9898\uff1a1. \u8003\u8651\u5230\u4e0d\u540c\u5e26\u5bbd\u548c\u5b58\u50a8\u7684\u9700\u6c42\uff0c\u5982\u4f55\u8bbe\u8ba1\u4e00\u4e2a\u80fd\u591f\u652f\u6301\u53ef\u53d8\u7801\u7387\u7684\u5355\u4e00\u6a21\u578b\uff0c\u4e14\u8be5\u7801\u7387\u8303\u56f4\u9700\u8981\u8db3\u591f\u5927\uff1b2. \u89c6\u9891\u7f16\u89e3\u7801\u5668\u901a\u8fc7\u6316\u6398\u65f6\u57df\u76f8\u5173\u6027\u4ee5\u51cf\u5c0f\u89c6\u9891\u5197\u4f59\uff0c\u4f46\u8fd9\u5f80\u5f80\u5bb9\u6613\u5bfc\u81f4\u65f6\u57df\u4e0a\u7684\u8bef\u5dee\u4f20\u64ad\u3002\u56e0\u6b64\uff0c\u5982\u4f55\u907f\u514d\u8be5\u8bef\u5dee\u4f20\u64ad\u9020\u6210\u7684\u91cd\u5efa\u8d28\u91cf\u8870\u51cf\u3002\u9488\u5bf9\u8fd9\u4e24\u4e2a\u95ee\u9898\uff0c\u7814\u7a76\u5458\u4eec\u901a\u8fc7\u8bbe\u8ba1\u7279\u5f81\u8c03\u5236\u6280\u672f\uff0c\u63d0\u51fa\u4e86\u76f8\u5e94\u7684\u89e3\u51b3\u65b9\u6848\u3002 \u4e3a\u4e86\u5b9e\u73b0\u5927\u8303\u56f4\u7684\u53ef\u53d8\u7801\u7387\uff0c\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4f7f\u7528\u53ef\u5b66\u4e60\u7684\u91cf\u5316\u7f29\u653e\u5668\u6765\u8c03\u6574\u6bcf\u5e27\u7684\u9690\u7279\u5f81\uff08\u5982\u56fe2\u53f3\u56fe\u6240\u793a\uff09\u3002\u5728\u8bad\u7ec3\u7684\u8fc7\u7a0b\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u8bbe\u8ba1\u4e86\u4e00\u79cd\u5747\u5300\u7684\u91cf\u5316\u53c2\u6570\u91c7\u6837\u673a\u5236\uff0c\u4ee5\u4f7f\u5f97\u7f16\u89e3\u7801\u5668\u5e94\u5bf9\u4e0d\u4e00\u6837\u7684\u7387\u5931\u771f\u6743\u8861\u3002 \u8be5\u8bad\u7ec3\u673a\u5236\u53ef\u4ee5\u5e2e\u52a9\u5b66\u4e60\u7ec6\u7c92\u5ea6\uff0c\u5e76\u4f18\u5316\u53ef\u8c03\u8282\u7684\u91cf\u5316\u7f29\u653e\u5668\uff0c\u4ece\u800c\u80fd\u5728\u4e00\u4e2a\u5f88\u5927\u7684\u7801\u7387\u8303\u56f4\u5185\uff0c\u6839\u636e\u7528\u6237\u9700\u6c42\u8c03\u8282\u89c6\u9891\u7f16\u7801\u8d28\u91cf\u2014\u2014\u53ef\u8c03\u7684\u89c6\u9891\u8d28\u91cf\u8303\u56f4\u5c06\u4ece\u4e4b\u524d\u76843.8 dB\u589e\u957f\u523011.4 dB\u3002\u53e6\u5916\uff0cDCVC-FM \u80fd\u591f\u52a8\u6001\u8c03\u8282\u89c6\u9891\u4e2d\u7684\u6bcf\u5e27\u8d28\u91cf\uff0c\u8be5\u529f\u80fd\u4e5f\u4f7f\u5f97\u6a21\u578b\u53ef\u4ee5\u5e94\u5bf9\u590d\u6742\u6613\u53d8\u7684\u7f51\u7edc\u5e26\u5bbd\u73af\u5883\uff0c\u56fe3\u4fbf\u5c55\u793a\u4e86 DCVC-FM \u5728\u5e94\u5bf9\u53d8\u5316\u5e26\u5bbd\u65b9\u9762\u7684\u80fd\u529b\u3002 \u9488\u5bf9\u65f6\u57df\u8bef\u5dee\u4f20\u64ad\u95ee\u9898\uff0c\u5df2\u6709\u7684\u6a21\u578b\u666e\u904d\u91c7\u7528\u9891\u7e41\u63d2\u5165\u9ad8\u8d28\u91cf\u5e27\u5185\u7f16\u7801\u5e27\u7684\u65b9\u6cd5\u6765\u5207\u65ad\u8bef\u5dee\u4f20\u64ad\u3002\u5c3d\u7ba1\u8be5\u65b9\u6cd5\u53ef\u4ee5\u6709\u6548\u904f\u5236\u8bef\u5dee\u7684\u6269\u6563\uff0c\u4f46\u7531\u4e8e\u5e27\u5185\u7f16\u7801\u7684\u6548\u7387\u6781\u4f4e\uff0c\u6574\u4f53\u7684\u89c6\u9891\u7f16\u7801\u6548\u7387\u5c06\u663e\u8457\u964d\u4f4e\u3002\u4e3a\u6b64\uff0c\u7814\u7a76\u5458\u4eec\u8bbe\u8ba1\u4e86\u4e00\u79cd\u7279\u5f81\u66f4\u65b0\u673a\u5236\u2014\u2014\u5f53\u76f8\u90bb\u5e27\u7684\u7279\u5f81\u5728\u65f6\u57df\u4f20\u64ad\u65f6\uff0c\u7814\u7a76\u5458\u4eec\u5c06\u5468\u671f\u6027\u5730\u4f7f\u7528\u8f85\u52a9\u6027\u7684\u7279\u5f81\u63d0\u53d6\u5668\u53bb\u8c03\u6574\u5e76\u66f4\u65b0\u88ab\u4f20\u64ad\u7684\u7279\u5f81\u3002\u8fd9\u4e00\u673a\u5236\u4f7f\u5f97 DCVC-FMDCVC-FM \u5728\u5904\u7406\u975e\u5e38\u957f\u7684\u89c6\u9891\u65f6\u4ecd\u80fd\u4fdd\u6301\u6781\u9ad8\u7684\u538b\u7f29\u6548\u7387\u3002\u5982\u56fe4\u6240\u793a\uff0cDCVC-FM \u5728\u538b\u7f29\u6027\u80fd\u65b9\u9762\u5df2\u7ecf\u8d85\u8fc7\u4e86\u6b63\u5728\u7814\u53d1\u7684\u4e0b\u4e00\u4ee3\u4f20\u7edf\u89c6\u9891\u7f16\u7801\u5668\u6807\u51c6\u7684\u539f\u578b ECM\u3002 \u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2311.10983 (opens in new tab) \u9879\u76ee\u94fe\u63a5\uff1ahttps:\/\/github.com\/XunshanMan\/MVGFormer (opens in new tab) [&hellip;]<\/p>\n","protected":false},"author":42735,"featured_media":1076694,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-content-parent":1012650,"footnotes":""},"research-area":[13556],"msr-locale":[268881],"msr-post-option":[],"class_list":["post-1073853","msr-blog-post","type-msr-blog-post","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-locale-zh_cn"],"msr_assoc_parent":{"id":1012650,"type":"lab"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1073853"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-blog-post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/42735"}],"version-history":[{"count":6,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1073853\/revisions"}],"predecessor-version":[{"id":1106160,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1073853\/revisions\/1106160"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1076694"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1073853"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1073853"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1073853"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1073853"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}