
{"id":10102,"date":"2026-07-10T13:07:22","date_gmt":"2026-07-10T05:07:22","guid":{"rendered":"https:\/\/infernews.com\/blog\/this-repository-contains-the-official-implementation-of-our-proposed-two-stage-f\/"},"modified":"2026-07-10T13:08:32","modified_gmt":"2026-07-10T05:08:32","slug":"this-repository-contains-the-official-implementation-of-our-proposed-two-stage-f","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/this-repository-contains-the-official-implementation-of-our-proposed-two-stage-f\/","title":{"rendered":"CrossViewTokenFusion\uff1a\u4e73\u623f X \u5149\u96d9\u8996\u89d2\u5206\u985e\u65b0\u8def\u7dda"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/pasted-934a0a1d5015.jpg\" alt=\"Repository image for PartAI-Projects\/CrossViewTokenFusion\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u81e8\u5e8a\u5224\u8b80\u4e73\u623f X \u5149\u5514\u6703\u53ea\u671b\u55ae\u4e00\u89d2\u5ea6\uff0c\u4f46\u4e0d\u5c11\u591a\u8996\u89d2\u65b9\u6cd5\u4ecd\u7136\u504f\u5411\u628a\u7279\u5fb5\u63d0\u65e9\u5408\u4f75\uff0c\u6216\u8005\u53ea\u505a\u4e00\u6b21 cross-attention\uff0c\u7d50\u679c\u5bb9\u6613\u628a view-specific \u7dda\u7d22\u540c\u5171\u4eab\u8cc7\u8a0a\u6df7\u57cb\u3002CrossViewTokenFusion \u5c6c\u65bc<strong>\u91ab\u5b78\u5f71\u50cf\u5206\u985e\u6a21\u578b\uff0f\u7814\u7a76\u539f\u578b<\/strong>\uff0c\u91dd\u5c0d dual-view mammography classification\uff0c\u91cd\u9ede\u4fc2\u8b93 CC \u8207 MLO \u5169\u500b\u8996\u89d2\u4ee5 token \u70ba\u55ae\u4f4d\u9010\u6b65\u4ea4\u63db\u8cc7\u8a0a\uff0c\u800c\u5514\u4fc2\u4e00\u958b\u59cb\u5c31\u7c97\u7565\u878d\u5408\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9805\u76ee\u5efa\u57fa\u65bc <a href=\"https:\/\/huggingface.co\/google\/medsiglip-448\" target=\"_blank\" rel=\"noreferrer noopener\">frozen MedSigLIP vision model<\/a>\uff0c\u63a1\u7528\u5169\u968e\u6bb5\u6d41\u7a0b\uff1a\u5148\u505a deep prompt learning \u9069\u914d\uff0c\u518d\u505a cross-view token-based fusion\u3002\u4f5c\u8005\u6279\u8a55\u65e2\u6709 multi-view learning \u5e38\u898b\u7684 feature-level aggregation \u540c single-stage cross-attention \u4e92\u52d5\u592a\u6dfa\uff0c\u65bc\u662f\u6539\u7528 dedicated fusion tokens \u4f5c\u70ba\u4e2d\u4ecb\uff0c\u900f\u904e cross-attention \u5728\u591a\u500b transformer \u6df1\u5ea6\u53cd\u8986\u50b3\u905e\u96d9\u5411\u8a0a\u606f\uff0c\u4e4b\u5f8c\u518d\u628a fusion tokens \u653e\u56de token sequence \u7e7c\u7e8c\u7d30\u5316\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u90e8\u7f72\u8981\u6c42\u5148\u6e96\u5099 VinDr-Mammo \u6216 CMMD\uff08Chinese Mammography Database\uff09\uff0c\u518d\u505a preprocessing\u3001stage 1 \u8a13\u7df4\u3001stage 2 \u8a13\u7df4\uff0c\u6700\u5f8c\u8f09\u5165 checkpoint \u6e2c\u8a66\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u6bd4\u8f03\u9069\u5408\u505a\u91ab\u5b78\u5f71\u50cf\u7814\u7a76\u3001\u91ab\u7642 AI \u5718\u968a\u9a57\u8b49\u591a\u8996\u89d2\u878d\u5408\u8a2d\u8a08\uff0c\u672a\u898b\u5230\u76f4\u63a5\u9762\u5411\u81e8\u5e8a\u7cfb\u7d71\u7684\u5c01\u88dd\u4ecb\u9762\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u820a\u7bc4\u5f0f\u591a\u6578\u7528 feature-level aggregation \u6216\u55ae\u5c64 fusion\uff1b\u9019\u500b\u9805\u76ee\u6539\u70ba multi-depth token \u4e92\u52d5<\/li>\n\n\n\n<li>\u4ee5 frozen vision transformer backbone \u914d\u5408 prompt learning\uff0c\u53d6\u5411\u4fc2\u5c11\u6539\u4e3b\u5e79\u3001\u591a\u505a\u9069\u914d<\/li>\n\n\n\n<li>\u516c\u958b\u8cc7\u6599\u96c6\u5305\u62ec VinDr-Mammo \u540c CMMD\uff0c\u65b9\u4fbf\u5b78\u8853\u91cd\u73fe\u8207\u6a6b\u5411\u6bd4\u8f03<\/li>\n\n\n\n<li>VinDr-Mammo BI-RADS classification \u9054\u5230 50.40% F1-score\u30010.8090 AUC<\/li>\n\n\n\n<li>\u76f8\u6bd4 dual-view fusion baseline\uff0c\u4e8c\u5206\u985e\u8a2d\u5b9a\u4e0b AUC \u63d0\u5347 0.10<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u7684\u50f9\u503c\u4fc2\u91cd\u65b0\u5b9a\u7fa9\u96d9\u8996\u89d2\u9ede\u6a23\u6e9d\u901a\uff1a\u4fdd\u7559\u5404\u81ea\u7d50\u69cb\uff0c\u518d\u7528 fusion tokens \u9010\u5c64\u50b3\u905e\u4e92\u88dc\u8b49\u64da\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/github.com\/PartAI-Projects\/CrossViewTokenFusion\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/arxiv.org\/pdf\/2607.06309\" rel=\"noopener noreferrer\"><strong>Paper<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u540c\u4e00\u500b\u75c5\u7076\u5728\u5169\u5f35\u4e73\u623f X \u5149\u89d2\u5ea6\u5165\u9762\uff0c\u7dda\u7d22\u672a\u5fc5\u4e00\u81f4\u3002CrossViewTokenFusion \u60f3\u8655\u7406\u7684\uff0c\u6b63\u6b63\u662f\u9ede\u6a23\u4fdd\u7559\u5169\u500b\u8996\u89d2\u5404\u81ea\u8a0a\u606f\uff0c\u518d\u9010\u5c64\u4ea4\u63db\u95dc\u9375\u8b49\u64da\u3002<\/p>\n","protected":false},"author":8,"featured_media":10101,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,144,119,76,127,149],"tags":[],"class_list":["post-10102","post","type-post","status-publish","format-standard","hentry","category-133","category-medical","category-119","category-76","category-127","category-149"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10102","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=10102"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10102\/revisions"}],"predecessor-version":[{"id":10104,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10102\/revisions\/10104"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10101"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10102"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10102"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10102"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}