{"id":1446,"date":"2025-04-09T00:04:14","date_gmt":"2025-04-08T16:04:14","guid":{"rendered":"https:\/\/cchsu.info\/wordpress\/?p=1446"},"modified":"2025-04-09T00:06:43","modified_gmt":"2025-04-08T16:06:43","slug":"we-just-won-top-2-6-244-performance-award-from-ntire-2025-reflection-removal-challenge-in-cvpr-2025","status":"publish","type":"post","link":"https:\/\/cchsu.info\/wordpress\/2025\/04\/09\/we-just-won-top-2-6-244-performance-award-from-ntire-2025-reflection-removal-challenge-in-cvpr-2025\/","title":{"rendered":"[Award] We just won Top 2% (6\/244) Performance Award from NTIRE 2025 Reflection Removal Challenge in CVPR 2025!"},"content":{"rendered":"
\n
\n
\n
\n
\n
\n
\n
\u518d\u6b21\u606d\u559c ACVLab \u6797\u5b87\u51e1<\/span><\/span><\/a><\/span> \u674e\u5bb6\u9298<\/span><\/span><\/a><\/span> \u653b\u514b IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2025 \u7684\u6700\u91cd\u8981\u7684 Low-Level Vision – NTIRE \u7af6\u8cfd\uff0c\u65bc Reflection Removal Challenge \u4e2d\uff0c\u7372\u5f97\u5168\u7403\u6392\u540d\u524d Top 2% \u7684\u540d\u6b21\uff0c\u4e5f\u662f\u53f0\u7063\u552f\u4e00\u9032\u5165\u524d Top 3% \u7684\u5718\u968a (6\/244)\u3002<\/div>\n<\/div>\n
\n
\u9019\u6b21\u53c3\u8207 Reflection Removal \u4efb\u52d9\uff0c\u6311\u6230\u6027\u6975\u9ad8\uff0c\u4e0d\u50c5\u8981\u6e96\u78ba\u53bb\u9664\u5f71\u50cf\u4e2d\u7169\u4eba\u7684\u53cd\u5c04\uff0c\u66f4\u8981\u5b8c\u6574\u91cd\u5efa\u88ab\u53cd\u5c04\u906e\u84cb\u7684\u9ad8\u983b\u7d30\u7bc0\u8207\u771f\u5be6\u7d0b\u7406\u8cc7\u8a0a\uff0c\u9019\u5728\u8a31\u591a\u5be6\u969b\u61c9\u7528\u4e2d\u90fd\u5177\u95dc\u9375\u50f9\u503c\uff0c\u4f8b\u5982\u81ea\u99d5\u8eca\u3001\u8996\u8a0a\u76e3\u63a7\u3001\u667a\u6167\u57ce\u5e02\u4ee5\u53ca\u5f71\u50cf\u589e\u5f37\u6280\u8853\u7b49\u9818\u57df\u3002\u70ba\u514b\u670d\u53cd\u5c04\u5340\u57df\u7684\u8cc7\u8a0a\u907a\u5931\u554f\u984c\uff0c\u5718\u968a\u5de7\u5999\u904b\u7528 OmniSR \u6a21\u578b\u4f5c\u70ba\u57fa\u790e\uff0c\u900f\u904e DINOv2 \u64f7\u53d6\u5f37\u5927\u7684\u8a9e\u610f\u7279\u5fb5\uff0c\u540c\u6642\u5229\u7528 Depth Anything V2 \u751f\u6210\u7cbe\u6e96\u7684\u6df1\u5ea6\u5716\uff0c\u4e26\u9032\u4e00\u6b65\u8f49\u63db\u70ba\u8868\u9762\u6cd5\u5411\u5716\uff0c\u4ee5\u63d0\u4f9b\u66f4\u8c50\u5bcc\u7684\u7d50\u69cb\u8cc7\u8a0a\u3002\u6b64\u5916\uff0c\u66f4\u7279\u5225\u5f15\u5165\u983b\u57df\u878d\u5408\u6a21\u7d44\uff08FreqFusion\uff09\uff0c\u6709\u6548\u6062\u5fa9\u56e0\u53cd\u5c04\u5f71\u97ff\u800c\u907a\u5931\u7684\u9ad8\u983b\u7d0b\u7406\uff0c\u7121\u9700\u984d\u5916\u5f8c\u8655\u7406\uff0c\u5373\u80fd\u5be6\u73fe\u9ad8\u54c1\u8cea\u7684\u7121\u53cd\u5c04\u5f71\u50cf\u91cd\u5efa\u3002<\/div>\n<\/div>\n
\n
\u6280\u8853\u5be6\u4f5c\u65b9\u9762\uff0c\u5718\u968a\u50c5\u4f7f\u7528\u7af6\u8cfd\u5b98\u65b9\u63d0\u4f9b\u7684\u8cc7\u6599\u96c6\u9032\u884c\u8a13\u7df4\uff0c\u900f\u904e\u5ba2\u88fd\u5316\u8a2d\u8a08\u7684\u5169\u968e\u6bb5\u8a13\u7df4\u7b56\u7565\uff0c\u5148\u4ee5 Charbonnier Loss \u7a69\u5065\u5c0d\u6297\u53cd\u5c04\u9020\u6210\u7684\u7570\u5e38\u503c (Robustness-aware training)\uff0c\u5f8c\u7e8c\u518d\u900f\u904e L2 Loss \u5fae\u8abf\u7d30\u7bc0\u8868\u73fe (Refinement-aware training)\uff0c\u6700\u7d42\u6210\u529f\u5728\u5ba2\u89c0\u54c1\u8cea\u8a55\u4f30\u4e0a\u9054\u5230\u5f88\u597d\u8868\u73fe\uff0c\u4e26\u5728\u5168\u7403\u9802\u5c16\u5718\u968a\u4e2d\u540d\u5217\u524d\u8305 (Top 2%)\u3002<\/div>\n<\/div>\n
\n
\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0c\u9019\u6b21\u5b87\u51e1\u518d\u6b21\u64d4\u7db1\u4e3b\u8981\u7814\u7a76\uff0c\u6301\u7e8c\u5c55\u73fe\u6975\u4f73\u7684\u89c0\u5bdf\u529b\u8207\u5275\u65b0\u80fd\u529b\uff0cACVLab \u65b0\u4e16\u4ee3\u7684\u5be6\u529b\u8207\u6f5b\u529b\u53ef\u671f\u554a\uff01\u606d\u559c\u5b87\u51e1\u3001\u5bb6\u9298\uff0c\u8d85\u5f37\u963f!<\/div>\n<\/div>\n
\n
—–<\/div>\n
More great news from ACVLab! Congratulations again to Yu-Fan Lin and Chia-Ming Lee for their remarkable achievement at the NTIRE Reflection Removal Challenge, held as part of the prestigious IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2025.<\/div>\n<\/div>\n
\n
Our team proudly ranked in the Top 2% worldwide (6th out of 244 teams), and we are also the only team from Taiwan to break into the Top 3%\u2014a truly outstanding accomplishment!<\/div>\n<\/div>\n
\n
Reflection removal is a highly challenging low-level vision task. It requires not only the accurate elimination of distracting reflections in images, but also the faithful reconstruction of high-frequency texture details and original surface structures occluded by those reflections. This technique holds immense practical value for applications like autonomous driving, video surveillance, smart city systems, and general image enhancement.<\/div>\n<\/div>\n
\n
To tackle this complex task, our team built upon the OmniSR architecture, leveraging semantic features extracted via DINOv2 and high-quality depth maps generated using Depth Anything V2. These depth maps were further converted into surface normal maps, enriching the model with structural cues. We also introduced a frequency-aware fusion module (FreqFusion) to restore high-frequency details lost due to reflections\u2014without the need for any post-processing. The result: high-resolution, reflection-free image reconstructions.<\/div>\n<\/div>\n
\n
Our training pipeline relied solely on the official competition dataset, following a two-stage training strategy. The first stage adopted Charbonnier Loss for its robustness against reflection-induced outliers, while the second stage fine-tuned texture fidelity using L2 Loss. The model consistently performed at a top-tier level in objective quality evaluations, securing our place among the global elite.<\/div>\n<\/div>\n
\n
A special shout-out to Yu-Fan, who once again led the core research effort with sharp insight and creativity. This achievement further proves the rising power and potential of our next-generation researchers at ACVLab.<\/div>\n<\/div>\n
\n
Major kudos to Yu-Fan and Chia-Ming\u2014amazing work! \"\ud83d\ude80\"<\/span><\/div>\n<\/div>\n
\n
[official website] https:\/\/codalab.lisn.upsaclay.fr\/competitions\/21394<\/a><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n
\n
\n
\n
\"\"<\/div>\n<\/div>\n
<\/div>\n<\/div>\n
<\/div>\n<\/div>\n
\n
\n
\n
\"\"<\/div>\n<\/div>\n
<\/div>\n<\/div>\n
<\/div>\n<\/div>\n
\n
\n
\n
\"\"<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n
\n
\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

\u518d\u6b21\u606d\u559c ACVLab \u6797\u5b87\u51e1 \u674e\u5bb6\u9298 \u653b\u514b IEEE\/CVF Conference on Computer […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[],"class_list":["post-1446","post","type-post","status-publish","format-standard","hentry","category-award"],"_links":{"self":[{"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/posts\/1446","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/comments?post=1446"}],"version-history":[{"count":2,"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/posts\/1446\/revisions"}],"predecessor-version":[{"id":1450,"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/posts\/1446\/revisions\/1450"}],"wp:attachment":[{"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/media?parent=1446"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/categories?post=1446"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cchsu.info\/wordpress\/wp-json\/wp\/v2\/tags?post=1446"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}