https://vilab.team/tag/detr/ DETR Hugo Blox Builder (https://hugoblox.com)en-usTue, 10 Mar 2026 00:00:00 +0000 https://vilab.team/media/icon_hu2896232876136423579.png https://vilab.team/tag/detr/ https://vilab.team/publication/rio-detr-detr-for-real-time-oriented-object-detection/ Tue, 10 Mar 2026 00:00:00 +0000 https://vilab.team/publication/rio-detr-detr-for-real-time-oriented-object-detection/ <p>本文提出RiO-DETR，一种面向实时旋转目标检测的DETR框架。针对方向语义依赖、角度周期性和搜索空间扩大等挑战，设计了内容驱动的角度估计、旋转校正正交注意力、解耦周期细化以及定向密集O2O机制，在保持实时效率的同时提升角度收敛速度与检测精度。在DOTA-1.0、DIOR-R和FAIR-1M-2.0上的实验表明，该方法实现了新的速度-精度平衡。</p> https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026-2/ Tue, 10 Mar 2026 00:00:00 +0000 https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026-2/ <p>热烈祝贺胡张驰同学！论文《RiO-DETR: DETR for Real-time Oriented Object Detection》已发表在 ECCV 2026 Oral。</p> <h2 id="rio-detr-detr-for-real-time-oriented-object-detection">RiO-DETR: DETR for Real-time Oriented Object Detection</h2> <p><strong>祝贺胡张驰同学！该论文已发表在 ECCV 2026。</strong></p> <p> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="RiO-DETR: DETR for Real-time Oriented Object Detection" srcset=" /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026-2/images/paper-01_hu3774487554520496472.webp 400w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026-2/images/paper-01_hu7272795672146682259.webp 760w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026-2/images/paper-01_hu2318532274489056445.webp 1200w" src="https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026-2/images/paper-01_hu3774487554520496472.webp" width="760" height="393" loading="lazy" data-zoomable /></div> </div></figure> </p> <ul> <li><strong>作者：</strong> Zhangchi Hu、Yifan Zhao、Yansong Peng、Wenzhang Sun、Xiangchen Yin、Jie Chen、Peixi Wu、Hebei Li、Xinghao Wang、Dongsheng Jiang、Xiaoyan Sun</li> <li><strong>发表载体：</strong> ECCV 2026</li> <li><strong>发表时间：</strong> 2026年3月10日</li> <li><strong>相关链接：</strong> <a href="https://arxiv.org/abs/2603.09411" target="_blank" rel="noopener">论文链接</a> · <a href="https://arxiv.org/pdf/2603.09411" target="_blank" rel="noopener">PDF</a> · <a href="https://github.com/RicePasteM/RiO-DETR" target="_blank" rel="noopener">代码</a></li> </ul> <h3 id="论文介绍">论文介绍</h3> <p>本文提出RiO-DETR，一种面向实时旋转目标检测的DETR框架。针对方向语义依赖、角度周期性和搜索空间扩大等挑战，设计了内容驱动的角度估计、旋转校正正交注意力、解耦周期细化以及定向密集O2O机制，在保持实时效率的同时提升角度收敛速度与检测精度。在DOTA-1.0、DIOR-R和FAIR-1M-2.0上的实验表明，该方法实现了新的速度-精度平衡。</p> https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026/ Tue, 10 Mar 2026 00:00:00 +0000 https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026/ <p>热烈祝贺胡张驰同学！论文《RiO-DETR: DETR for Real-time Oriented Object Detection》已发表在 ECCV 2026。</p> <h2 id="rio-detr-detr-for-real-time-oriented-object-detection">RiO-DETR: DETR for Real-time Oriented Object Detection</h2> <p><strong>祝贺胡张驰同学！该论文已发表在 ECCV 2026。</strong></p> <p> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="RiO-DETR: DETR for Real-time Oriented Object Detection" srcset=" /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026/images/paper-01_hu3774487554520496472.webp 400w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026/images/paper-01_hu7272795672146682259.webp 760w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026/images/paper-01_hu2318532274489056445.webp 1200w" src="https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8%E4%BA%8E-eccv-2026/images/paper-01_hu3774487554520496472.webp" width="760" height="393" loading="lazy" data-zoomable /></div> </div></figure> </p> <ul> <li><strong>作者：</strong> Zhangchi Hu、Yifan Zhao、Yansong Peng、Wenzhang Sun、Xiangchen Yin、Jie Chen、Peixi Wu、Hebei Li、Xinghao Wang、Dongsheng Jiang、Xiaoyan Sun</li> <li><strong>发表载体：</strong> ECCV 2026</li> <li><strong>发表时间：</strong> 2026年3月10日</li> <li><strong>相关链接：</strong> <a href="https://arxiv.org/abs/2603.09411" target="_blank" rel="noopener">论文链接</a> · <a href="https://arxiv.org/pdf/2603.09411" target="_blank" rel="noopener">PDF</a> · <a href="https://github.com/RicePasteM/RiO-DETR" target="_blank" rel="noopener">代码</a></li> </ul> <h3 id="论文介绍">论文介绍</h3> <p>本文提出RiO-DETR，一种面向实时旋转目标检测的DETR框架。针对方向语义依赖、角度周期性和搜索空间扩大等挑战，设计了内容驱动的角度估计、旋转校正正交注意力、解耦周期细化以及定向密集O2O机制，在保持实时效率的同时提升角度收敛速度与检测精度。在DOTA-1.0、DIOR-R和FAIR-1M-2.0上的实验表明，该方法实现了新的速度-精度平衡。</p> https://vilab.team/publication/dome-detr-detr-with-density-oriented-feature-query-manipulat/ Mon, 27 Oct 2025 00:00:00 +0000 https://vilab.team/publication/dome-detr-detr-with-density-oriented-feature-query-manipulat/ <p>本文提出 Dome-DETR，一种面向微小物体检测的高效框架。针对现有方法特征利用不充分和计算成本高的问题，引入轻量级密度聚焦提取器（DeFE）生成紧凑前景掩码，并基于掩码的窗口注意力稀疏化（MWAS）将计算资源集中于关键区域。同时提出渐进自适应查询初始化（PAQI），自适应调节查询分布，提升检测效率与精度。</p> https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-3-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8/ Mon, 27 Oct 2025 00:00:00 +0000 https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-3-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8/ <p>热烈祝贺第一作者 Changsheng Gao、胡张驰同学、王义钧同学！近期，实验室共有 3 项科研成果正式发表在ACM MM 2025。</p> <h2 id="dt-ufc-universal-large-model-feature-coding-via-peaky-to-balanced-distribution-transformation">DT-UFC: Universal large model feature coding via peaky-to-balanced distribution transformation</h2> <p><strong>祝贺第一作者 Changsheng Gao！该论文已发表在 <em>ACM MM</em> 2025。</strong></p> <ul> <li><strong>作者：</strong> Changsheng Gao、Zijie Liu、Li Li、Dong Liu、Xiaoyan Sun、Weisi Lin</li> <li><strong>发表载体：</strong> <em>ACM MM</em> 2025</li> <li><strong>发表时间：</strong> 2025年10月27日</li> <li><strong>相关链接：</strong> <a href="https://dl.acm.org/doi/abs/10.1145/3746027.3755814" target="_blank" rel="noopener">论文链接</a> · <a href="https://dl.acm.org/doi/pdf/10.1145/3746027.3755814" target="_blank" rel="noopener">PDF</a></li> </ul> <h3 id="论文介绍">论文介绍</h3> <p>本文首次系统研究大模型通用特征编码问题，旨在降低分布式部署中的传输与存储开销。针对不同模型提取特征分布差异显著（如DINOv2特征高度集中、SD3特征分散均匀）导致压缩效率低、跨模型泛化难的问题，提出一种从峰值分布到平衡分布的变换方法，实现通用且高效的特征压缩与传输。</p> <hr> <h2 id="dome-detr-detr-with-density-oriented-feature-query-manipulation-for-efficient-tiny-object-detection">Dome-DETR: DETR with density-oriented feature-query manipulation for efficient tiny object detection</h2> <p><strong>祝贺胡张驰同学！该论文已发表在 <em>ACM MM</em>。</strong></p> <ul> <li><strong>作者：</strong> Zhangchi Hu、Peixi Wu、Jie Chen、Huyue Zhu、Yijun Wang、Yansong Peng、Hebei Li、Xiaoyan Sun</li> <li><strong>发表载体：</strong> <em>ACM MM</em></li> <li><strong>发表时间：</strong> 2025年10月27日</li> <li><strong>相关链接：</strong> <a href="https://dl.acm.org/doi/abs/10.1145/3746027.3754861" target="_blank" rel="noopener">论文链接</a> · <a href="https://dl.acm.org/doi/pdf/10.1145/3746027.3754861" target="_blank" rel="noopener">PDF</a> · <a href="https://github.com/RicePasteM/Dome-DETR" target="_blank" rel="noopener">代码</a></li> </ul> <h3 id="论文介绍-1">论文介绍</h3> <p>本文提出 Dome-DETR，一种面向微小物体检测的高效框架。针对现有方法特征利用不充分和计算成本高的问题，引入轻量级密度聚焦提取器（DeFE）生成紧凑前景掩码，并基于掩码的窗口注意力稀疏化（MWAS）将计算资源集中于关键区域。同时提出渐进自适应查询初始化（PAQI），自适应调节查询分布，提升检测效率与精度。</p> <hr> <h2 id="medkcoop-dual-knowledge-guided-graph-prompt-learning-for-biomedical-vision-language-models">MeDKCoOp: Dual Knowledge-guided Graph Prompt Learning for Biomedical Vision-Language Models</h2> <p><strong>祝贺王义钧同学！该论文已发表在 <em>ACM MM</em>。</strong></p> <ul> <li><strong>作者：</strong> Yijun Wang、Siying Wu、Lubin Gan、Zheyu Zhang、Jing Zhang、Zhangchi Hu、Huyue Zhu、Peixi Wu、Xiaoyan Sun</li> <li><strong>发表载体：</strong> <em>ACM MM</em></li> <li><strong>发表时间：</strong> 2025年10月27日</li> <li><strong>相关链接：</strong> <a href="https://dl.acm.org/doi/abs/10.1145/3746027.3755135" target="_blank" rel="noopener">论文链接</a> · <a href="https://dl.acm.org/doi/pdf/10.1145/3746027.3755135" target="_blank" rel="noopener">PDF</a></li> </ul> <h3 id="论文介绍-2">论文介绍</h3> <p>本文提出MeDKCoOp，一种面向生物医学视觉语言模型的双知识引导图提示学习方法。该方法系统整合医学领域知识，从文本与视觉分支提取专门知识并构建图结构表示，通过知识引导的关系转移实现跨模态融合，并动态优化可学习提示，以增强CLIP等模型在医学下游任务中的适应能力。实验表明其在多个生物医学基准上取得优异性能。</p> https://vilab.team/publication/d-fine-redefine-regression-task-of-detrs-as-fine-grained-dis/ Thu, 01 May 2025 00:00:00 +0000 https://vilab.team/publication/d-fine-redefine-regression-task-of-detrs-as-fine-grained-dis/ <p>D-FINE是一种实时目标检测器，通过重新定义DETR中的边界框回归任务实现高精度定位。其核心包含细粒度分布细化（FDR）和全局最优定位自蒸馏（GO-LSD）。FDR将回归从预测固定坐标改为迭代细化概率分布，提供细粒度中间表示；GO-LSD通过自蒸馏将定位知识从最终层传递到浅层，并简化深层残差预测。在COCO上达到54.0%/55.8% AP，124/78 FPS，预训练后达57.1%/59.3% AP，超越现有实时检测器，并显著提升多种DETR模型性能。</p> https://vilab.team/event/publication-news-06e581b9272339ee/ Thu, 01 May 2025 00:00:00 +0000 https://vilab.team/event/publication-news-06e581b9272339ee/ <p>热烈祝贺彭岩松同学！论文《D-FINE: Redefine regression task of DETRs as fine-grained distribution refinement》已发表在 <em>ICLR 2025</em>。</p> <h2 id="d-fine-redefine-regression-task-of-detrs-as-fine-grained-distribution-refinement">D-FINE: Redefine regression task of DETRs as fine-grained distribution refinement</h2> <p><strong>祝贺彭岩松同学！该论文已发表在 <em>ICLR</em>。</strong></p> <p> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="D-FINE: Redefine regression task of DETRs as fine-grained distribution refinement" srcset=" /event/publication-news-06e581b9272339ee/images/paper-01_hu14730101056268865597.webp 400w, /event/publication-news-06e581b9272339ee/images/paper-01_hu1666356483600992880.webp 760w, /event/publication-news-06e581b9272339ee/images/paper-01_hu15779176642512148333.webp 1200w" src="https://vilab.team/event/publication-news-06e581b9272339ee/images/paper-01_hu14730101056268865597.webp" width="760" height="314" loading="lazy" data-zoomable /></div> </div></figure> </p> <ul> <li><strong>作者：</strong> Yansong Peng、Hebei Li、Peixi Wu、Yueyi Zhang、Xiaoyan Sun、Feng Wu</li> <li><strong>发表载体：</strong> <em>ICLR</em></li> <li><strong>发表时间：</strong> 2025年5月1日</li> <li><strong>相关链接：</strong> <a href="https://proceedings.iclr.cc/paper_files/paper/2025/hash/6cf58a87e3097e7d1f9be3e8693a93de-Abstract-Conference.html" target="_blank" rel="noopener">论文链接</a> · <a href="https://proceedings.iclr.cc/paper_files/paper/2025/file/6cf58a87e3097e7d1f9be3e8693a93de-Paper-Conference.pdf" target="_blank" rel="noopener">PDF</a> · <a href="https://github.com/Peterande/D-FINE" target="_blank" rel="noopener">代码</a></li> </ul> <h3 id="论文介绍">论文介绍</h3> <p>D-FINE是一种实时目标检测器，通过重新定义DETR中的边界框回归任务实现高精度定位。其核心包含细粒度分布细化（FDR）和全局最优定位自蒸馏（GO-LSD）。FDR将回归从预测固定坐标改为迭代细化概率分布，提供细粒度中间表示；GO-LSD通过自蒸馏将定位知识从最终层传递到浅层，并简化深层残差预测。在COCO上达到54.0%/55.8% AP，124/78 FPS，预训练后达57.1%/59.3% AP，超越现有实时检测器，并显著提升多种DETR模型性能。</p>