https://vilab.team/tag/%E9%AB%98%E6%96%AF%E6%BA%85%E5%B0%84/ 高斯溅射 Hugo Blox Builder (https://hugoblox.com)en-usSun, 19 Oct 2025 00:00:00 +0000 https://vilab.team/media/icon_hu2896232876136423579.png https://vilab.team/tag/%E9%AB%98%E6%96%AF%E6%BA%85%E5%B0%84/ https://vilab.team/publication/dash-4d-hash-encoding-with-self-supervised-decomposition-for/ Sun, 19 Oct 2025 00:00:00 +0000 https://vilab.team/publication/dash-4d-hash-encoding-with-self-supervised-decomposition-for/ <p>本文提出DASH，一种实时动态场景渲染框架，采用4D哈希编码结合自监督分解。针对现有平面基动态高斯溅射方法因低秩假设导致特征重叠和渲染质量差的问题，DASH通过自监督分解机制分离动态与静态组件，无需人工标注或预计算掩码，并引入多分辨率4D哈希编码器对动态元素进行显式表示，避免低秩约束，从而减少哈希冲突和冗余，实现高质量实时渲染。</p> https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/ Sun, 19 Oct 2025 00:00:00 +0000 https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/ <p>热烈祝贺陈杰同学、吴沛熹同学！近期，实验室共有 2 项科研成果正式发表在ICCV 2025。</p> <h2 id="dash-4d-hash-encoding-with-self-supervised-decomposition-for-real-time-dynamic-scene-rendering">Dash: 4d hash encoding with self-supervised decomposition for real-time dynamic scene rendering</h2> <p><strong>祝贺陈杰同学！该论文已发表在 <em>ICCV</em>。</strong></p> <p> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="Dash: 4d hash encoding with self-supervised decomposition for real-time dynamic scene rendering" srcset=" /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-01_hu14668970062868369588.webp 400w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-01_hu9539442214703728986.webp 760w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-01_hu17969550026126564787.webp 1200w" src="https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-01_hu14668970062868369588.webp" width="760" height="339" loading="lazy" data-zoomable /></div> </div></figure> </p> <ul> <li><strong>作者：</strong> Jie Chen、Zhangchi Hu、Peixi Wu、Huyue Zhu、Hebei Li、Xiaoyan Sun</li> <li><strong>发表载体：</strong> <em>ICCV</em></li> <li><strong>发表时间：</strong> 2025年10月19日</li> <li><strong>相关链接：</strong> <a href="https://ieeexplore.ieee.org/abstract/document/11444039/" target="_blank" rel="noopener">论文链接</a> · <a href="https://arxiv.org/pdf/2507.19141" target="_blank" rel="noopener">PDF</a> · <a href="https://github.com/chenj02/DASH" 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DASH，一种实时动态场景渲染框架，采用4D哈希编码结合自监督分解。针对现有平面基动态高斯溅射方法因低秩假设导致特征重叠和渲染质量差的问题，DASH通过自监督分解机制分离动态与静态组件，无需人工标注或预计算掩码，并引入多分辨率4D哈希编码器对动态元素进行显式表示，避免低秩约束，从而减少哈希冲突和冗余，实现高质量实时渲染。</p> <hr> <h2 id="efficient-spiking-point-mamba-for-point-cloud-analysis">Efficient spiking point mamba for point cloud analysis</h2> <p><strong>祝贺吴沛熹同学！该论文已发表在 <em>ICCV</em>。</strong></p> <p> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="Efficient spiking point mamba for point cloud analysis" srcset=" /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-02_hu18026702173288992064.webp 400w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-02_hu8515574002487600796.webp 760w, /event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-02_hu2828688134925209604.webp 1200w" src="https://vilab.team/event/%E7%A5%9D%E8%B4%BA%E5%AE%9E%E9%AA%8C%E5%AE%A4-2-%E9%A1%B9%E7%A7%91%E7%A0%94%E6%88%90%E6%9E%9C%E5%8F%91%E8%A1%A8-2/images/paper-02_hu18026702173288992064.webp" width="760" height="419" loading="lazy" data-zoomable /></div> </div></figure> </p> <ul> <li><strong>作者：</strong> Peixi Wu、Bosong Chai、Menghua Zheng、Wei Li、Zhangchi Hu、Jie Chen、Zheyu Zhang、Hebei Li、Xiaoyan Sun</li> <li><strong>发表载体：</strong> <em>ICCV</em></li> <li><strong>发表时间：</strong> 2025年10月19日</li> <li><strong>相关链接：</strong> <a href="https://ieeexplore.ieee.org/abstract/document/11445417/" target="_blank" rel="noopener">论文链接</a> · <a href="https://arxiv.org/pdf/2504.14371" target="_blank" rel="noopener">PDF</a> · <a href="https://github.com/PeppaWu/SPM" 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 Spiking Point Mamba (SPM)，这是首个将 Mamba 引入三维点云分析的脉冲神经网络。针对直接适配 Mamba 时存在的时序动态不匹配和脉冲引起的信息损失问题，作者设计了层次动态编码 (HDE) 以增强动态时序建模，并提出 Spiking Mamba Block (SMB) 来学习跨时间步特征并减少脉冲信息丢失。此外，采用非对称 SNN 训练策略进一步提升性能。SPM 可作为高效骨干网络，适用于点云分类、部件分割与重建等任务。</p>