https://rubatotree.github.io/blog/tags/3d-gaussians/Recent content in 3D Gaussians on Umbrella Coffeehttps://rubatotree.github.io/blog/images/og-default.pnghttps://rubatotree.github.io/blog/images/og-default.pngHugoen-usThu, 04 Jun 2026 00:00:00 +0800https://rubatotree.github.io/blog/posts/embodied-notes-1-robosplat/Thu, 04 Jun 2026 00:00:00 +0800https://rubatotree.github.io/blog/posts/embodied-notes-1-robosplat/<p>这周组会上导师提到现在做 Rendering 已经很难有价值和影响力了，也很难找到好的课题，应该 all in 具身智能。感觉自己从一个转型阵痛期的组跳进了另一个转型阵痛期的组。不过转型是正确的，具身方向的确好发有影响力的论文，就业前景也比传统图形学好。虽然私心比较喜欢简洁优雅但没用的图形学，但确实应该现实一点。我们该思考一下怎样把已有的图形学、渲染经验搬到具身智能的应用上。</p> <p>组里同学分享的 RoboSplat <a id="loc-1" href="#loc-10" role="doc-biblioref">[1]</a> 一文确实相当契合 CG for embodied 的愿景，因此我们从这篇论文开始了解我们能够做什么。</p> <h2 id="loc-2">论文概要</h2> <p>本文的目标是通过输入场景的多视角 RGB 图片和<strong>单次</strong>示教样本（Expert Demonstration，人工操作机械臂完成一次任务，得到每帧的机械臂状态数据和单监督视角 RGB 图像序列）训练 VLA 模型泛化完成不同干扰场景下的同一任务。本文试图解决的干扰场景分为物体位姿移动(Object Pose)、监督相机移动(Camera View)、光照条件改变(Lighting)、物体种类改变(Object Type)、周边环境改变(Appearance)、机械臂外观改变(Cross Embodiment)六种。</p> <p class="typst-parbreak"></p> <div style="display: grid; place-items: start center;"> <figure> <div style="display: grid; place-items: start center;"><img src="https://rubatotree.github.io/blog/images/embodied-notes-1-robosplat/robosplat-teaser.png" alt loading="lazy"></div> <div style="display: grid; place-items: start center;"> <figcaption>图 1 论文 teaser，展示了 RoboSplat 的训练过程。</figcaption> </div> </figure> </div> <p class="typst-parbreak"></p> <p>本文做到泛化的具体方法则是做<strong>数据增强</strong>，从单次示教样本中根据这几种干扰情况生成上千条增强样本 (augmented demonstrations)。以往生成增强样本的思路是直接在 2D 图像空间做生成式编辑，而本文考虑将多视角图片训练成 3DGS 场景（并在语义上分离场景、物体、机械臂，给机械臂绑骨），去在 3DGS 场景中做编辑模拟这些干扰，再合成回样本图像序列。这样做的好处是保证了图像之间场景的一致性，且不会有抖动等情况出现，因此能准确指导 VLA 模型学习到正确的操作策略。</p> <p>本文对这几种任务做了测试：</p> <p></p> <ul> <li> <p>Pick Object（抓取物体）：机械臂需要抓起放置在 30cm*40cm 范围桌面上的目标物体。</p> <p></p> <ul> <li>难点： 测试基本的空间 3D 定位与抓取能力。</li> </ul> <p></p> </li> <li> <p>Close Drawer（关闭抽屉）：机械臂需要推关一个抽屉。抽屉的位置在 15cm*40cm 范围内随机变化，且抽屉在 Z 轴方向的旋转角度在 [-pi/8, pi/8] 之间随机偏转。</p>https://rubatotree.github.io/blog/posts/sig26-paper-notes-1/Wed, 27 May 2026 00:00:00 +0800https://rubatotree.github.io/blog/posts/sig26-paper-notes-1/<img src="https://rubatotree.github.io/blog/images/sig26-paper-notes/teaser-1.png" alt loading="lazy"> <p>写一些笔记记录一下领域的最新进展。会持续更新，主页的 blog 发表日期仅供参考~</p> <h2 id="loc-1">1. 正向渲染加速</h2> <h3 id="loc-2">Gaussian Point Splatting <a id="loc-3" href="#loc-24" role="doc-biblioref">[1]</a> [<a href="https://jorisar.nl/gaussian_point_splatting/" target="_blank" rel="noopener noreferrer"><span style="color: #59a4ff;"><span style="text-decoration: underline">Project</span></span></a>]</h3> <img src="https://rubatotree.github.io/blog/images/sig26-paper-notes/Rijsdijk2026GaussianPointSplatting.png" alt loading="lazy"> <p>本文用蒙特卡洛方法代替深度排序，加速了超大 3DGS 场景在 GPU 上的正向渲染过程。目前仅支持正向渲染。</p> <p>本文注意到单个 Gaussian Splat 到屏幕空间的行为可以等价于在屏幕空间上按一个 Gaussian 分布放回采样 <span role="math"><svg class="typst-frame" style="overflow: visible; width: 0.909em; height: 0.683em;" viewBox="0 0 12.271500000000001 9.2205" width="12.271500000000001pt" height="9.2205pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:h5="http://www.w3.org/1999/xhtml"><g><g class="typst-text" transform="matrix(1 0 0 -1 0 9.2205)"><use xlink:href="#g97EE8E40EE0EEE4FD2AF6896DC5BA0D7" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g></g><defs id="glyph"><symbol id="g97EE8E40EE0EEE4FD2AF6896DC5BA0D7" overflow="visible"><path d="M 0 0m 11.8935 9.0720005 c 0 0 0 0.14849949 -0.17549992 0.14849949 c -0.44550037 0 -0.9180002 -0.040499687 -1.3635006 -0.040499687 c -0.45899963 0 -0.9314995 0.040499687 -1.3769999 0.040499687 c -0.080999374 0 -0.24300003 0 -0.24300003 -0.2699995 c 0 -0.14850044 0.13500023 -0.14850044 0.24300003 -0.14850044 c 0.76950073 -0.013500214 0.9180002 -0.29699993 0.9180002 -0.59399986 c 0 -0.040499687 -0.026999474 -0.24300003 -0.040499687 -0.2835002 l -1.5120001 -6.0074997 l -2.9835005 7.047001 c -0.1079998 0.24299908 -0.121500015 0.2564993 -0.43199968 0.2564993 h -1.8090003 c -0.26999998 0 -0.3915 0 -0.3915 -0.2699995 c 0 -0.14850044 0.121500015 -0.14850044 0.37800002 -0.14850044 c 0.067500114 0 0.9180002 0 0.9180002 -0.121500015 l -1.8090003 -7.2495003 c -0.13499999 -0.53999996 -0.36449993 -0.972 -1.458 -1.0125 c -0.08100003 0 -0.22950006 -0.013500005 -0.22950006 -0.26999998 c 0 -0.094500005 0.067500055 -0.14850001 0.17550004 -0.14850001 c 0.43200004 0 0.9045 0.0405 1.35 0.0405 c 0.4590001 0 0.94499993 -0.0405 1.3905001 -0.0405 c 0.067499876 0 0.24300003 0 0.24300003 0.27 c 0 0.13499999 -0.121500015 0.1485 -0.26999998 0.1485 c -0.783 0.02700001 -0.89100003 0.324 -0.89100003 0.59400004 c 0 0.094499946 0.013499975 0.16199994 0.0539999 0.31050003 l 1.7820003 7.1280003 c 0.0539999 -0.08100033 0.0539999 -0.1079998 0.121500015 -0.24300003 l 3.3614998 -7.9515 c 0.094500065 -0.22950001 0.13500023 -0.2565 0.25650024 -0.2565 c 0.14849949 0 0.14849949 0.0405 0.2159996 0.28350002 l 1.8900003 7.5195 c 0.13500023 0.54000044 0.37800026 0.9585004 1.4580002 0.9990001 c 0.067500114 0 0.22949982 0.013500214 0.22949982 0.27000046 Z "/></symbol></defs></svg></span> 个点的结果的期望（<span role="math"><svg class="typst-frame" style="overflow: visible; width: 0.909em; height: 0.683em;" viewBox="0 0 12.271500000000001 9.2205" width="12.271500000000001pt" height="9.2205pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:h5="http://www.w3.org/1999/xhtml"><g><g class="typst-text" transform="matrix(1 0 0 -1 0 9.2205)"><use xlink:href="#g97EE8E40EE0EEE4FD2AF6896DC5BA0D7" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g></g><defs id="glyph"><symbol id="g97EE8E40EE0EEE4FD2AF6896DC5BA0D7" overflow="visible"><path d="M 0 0m 11.8935 9.0720005 c 0 0 0 0.14849949 -0.17549992 0.14849949 c -0.44550037 0 -0.9180002 -0.040499687 -1.3635006 -0.040499687 c -0.45899963 0 -0.9314995 0.040499687 -1.3769999 0.040499687 c -0.080999374 0 -0.24300003 0 -0.24300003 -0.2699995 c 0 -0.14850044 0.13500023 -0.14850044 0.24300003 -0.14850044 c 0.76950073 -0.013500214 0.9180002 -0.29699993 0.9180002 -0.59399986 c 0 -0.040499687 -0.026999474 -0.24300003 -0.040499687 -0.2835002 l -1.5120001 -6.0074997 l -2.9835005 7.047001 c -0.1079998 0.24299908 -0.121500015 0.2564993 -0.43199968 0.2564993 h -1.8090003 c -0.26999998 0 -0.3915 0 -0.3915 -0.2699995 c 0 -0.14850044 0.121500015 -0.14850044 0.37800002 -0.14850044 c 0.067500114 0 0.9180002 0 0.9180002 -0.121500015 l -1.8090003 -7.2495003 c -0.13499999 -0.53999996 -0.36449993 -0.972 -1.458 -1.0125 c -0.08100003 0 -0.22950006 -0.013500005 -0.22950006 -0.26999998 c 0 -0.094500005 0.067500055 -0.14850001 0.17550004 -0.14850001 c 0.43200004 0 0.9045 0.0405 1.35 0.0405 c 0.4590001 0 0.94499993 -0.0405 1.3905001 -0.0405 c 0.067499876 0 0.24300003 0 0.24300003 0.27 c 0 0.13499999 -0.121500015 0.1485 -0.26999998 0.1485 c -0.783 0.02700001 -0.89100003 0.324 -0.89100003 0.59400004 c 0 0.094499946 0.013499975 0.16199994 0.0539999 0.31050003 l 1.7820003 7.1280003 c 0.0539999 -0.08100033 0.0539999 -0.1079998 0.121500015 -0.24300003 l 3.3614998 -7.9515 c 0.094500065 -0.22950001 0.13500023 -0.2565 0.25650024 -0.2565 c 0.14849949 0 0.14849949 0.0405 0.2159996 0.28350002 l 1.8900003 7.5195 c 0.13500023 0.54000044 0.37800026 0.9585004 1.4580002 0.9990001 c 0.067500114 0 0.22949982 0.013500214 0.22949982 0.27000046 Z "/></symbol></defs></svg></span> 和 Gaussian 的参数相关）。而一系列 Gaussians 做 Alpha Blending 的行为可以等价于，对每个 Gaussian <span role="math"><svg class="typst-frame" style="overflow: visible; width: 1.0248em; height: 0.683em;" viewBox="0 0 13.8348 9.2205" width="13.8348pt" height="9.2205pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:h5="http://www.w3.org/1999/xhtml"><g><g class="typst-text" transform="matrix(1 0 0 -1 0 9.2205)"><use xlink:href="#g16FCB66E7EC3CC47C0739BCD672A0DEE" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 9.261 12.555)"><use xlink:href="#g3A69B5BA14C4FC0AE0508775E710B241" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g></g><defs id="glyph"><symbol id="g16FCB66E7EC3CC47C0739BCD672A0DEE" overflow="visible"><path d="M 0 0m 10.26 9.382501 c 0 0.040499687 -0.027000427 0.13499928 -0.14850044 0.13499928 c -0.040499687 0 -0.0539999 -0.01349926 -0.20249939 -0.1619997 l -0.9449997 -1.0395002 c -0.121500015 0.18900013 -0.7425003 1.2014999 -2.2410007 1.2014999 c -3.0105 0 -6.048 -2.9834995 -6.048 -6.1154995 c 0 -2.1465 1.4985001 -3.6990001 3.6855004 -3.6990001 c 0.59399986 0 1.2014999 0.121500015 1.6875 0.32400003 c 0.6749997 0.27 0.93149996 0.5535 1.1745 0.82350004 c 0.121500015 -0.33750004 0.47249985 -0.8370001 0.6075001 -0.8370001 c 0.06749964 0 0.09449959 0.054 0.09449959 0.054 c 0.027000427 0.027000003 0.16200018 0.54 0.2295003 0.82350004 l 0.25650024 1.0395 c 0.0539999 0.22950006 0.121500015 0.4590001 0.17549992 0.6884999 c 0.14849949 0.6075001 0.1619997 0.6345 0.9314995 0.64800024 c 0.067500114 0 0.21600056 0.013499975 0.21600056 0.26999998 c 0 0.09449983 -0.067500114 0.14849997 -0.17549992 0.14849997 c -0.31050014 0 -1.1070004 -0.040500164 -1.4175005 -0.040500164 l -1.8899999 0.040500164 c -0.121500015 0 -0.2835002 0 -0.2835002 -0.26999998 c 0 -0.14849997 0.10800028 -0.14849997 0.4050002 -0.14849997 c 0 0 0.40499973 0 0.7154999 -0.027000189 c 0.3510003 -0.040499926 0.41849995 -0.08099985 0.41849995 -0.2565 c 0 -0.121500015 -0.14849997 -0.72899985 -0.28349972 -1.2285 c -0.37800026 -1.485 -2.1330004 -1.6335 -2.6055002 -1.6335 c -1.296 0 -2.7135 0.7695 -2.7135 2.835 c 0 0.41849995 0.13499999 2.646 1.5525 4.401 c 0.72899985 0.9180002 2.0385 1.7414999 3.3750002 1.7414999 c 1.3769999 0 2.1735 -1.0395002 2.1735 -2.6054997 c 0 -0.53999996 -0.040499687 -0.5535002 -0.040499687 -0.6884999 c 0 -0.13500023 0.14849949 -0.13500023 0.20249939 -0.13500023 c 0.17549992 0 0.17549992 0.02699995 0.24300003 0.26999998 Z "/></symbol><symbol id="g3A69B5BA14C4FC0AE0508775E710B241" overflow="visible"><path d="M 0 0m 3.0618 5.90625 c 0 0.15119982 -0.11339998 0.35909986 -0.37800002 0.35909986 c -0.25515008 0 -0.5292001 -0.24569988 -0.5292001 -0.5197501 c 0 -0.16064978 0.12284994 -0.35909986 0.37800002 -0.35909986 c 0.27405 0 0.5292001 0.2645998 0.5292001 0.5197501 Z m 0.24569988 -4.5549 c 0 0.12285006 -0.12284994 0.12285006 -0.15120006 0.12285006 c -0.1322999 0 -0.14174986 -0.05669999 -0.17954993 -0.16065001 c -0.21735 -0.756 -0.63314986 -1.14345 -1.0016999 -1.14345 c -0.18900001 0 -0.23625004 0.12285 -0.23625004 0.33074996 c 0 0.21735 0.06615007 0.3969 0.15120006 0.6048 l 0.3024 0.75600004 l 0.4630499 1.20015 c 0.028350115 0.09449983 0.05669999 0.20789981 0.05669999 0.30239987 c 0 0.44414997 -0.37800002 0.8032501 -0.8977499 0.8032501 c -0.93555 0 -1.37025 -1.2852001 -1.37025 -1.4458499 c 0 -0.12285018 0.13229999 -0.12285018 0.16064999 -0.12285018 c 0.13230002 0 0.14174998 0.047250032 0.17009997 0.15120006 c 0.2457 0.81270003 0.6615 1.1529 1.01115 1.1529 c 0.15119994 0 0.23625004 -0.07559991 0.23625004 -0.33075 c 0 -0.21735 -0.05669999 -0.3591001 -0.29295003 -0.94499993 l -0.59535 -1.52145 c -0.037799954 -0.12284994 -0.08504999 -0.23624998 -0.08504999 -0.39689994 c 0 -0.44415003 0.37800002 -0.80325 0.89775 -0.80325 c 0.94500005 0 1.3607999 1.3040999 1.3607999 1.4458499 Z "/></symbol></defs></svg></span> 采样 <span role="math"><svg class="typst-frame" style="overflow: visible; width: 1.1018000000000001em; height: 0.683em;" viewBox="0 0 14.8743 9.2205" width="14.8743pt" height="9.2205pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:h5="http://www.w3.org/1999/xhtml"><g><g class="typst-text" transform="matrix(1 0 0 -1 0 9.2205)"><use xlink:href="#g97EE8E40EE0EEE4FD2AF6896DC5BA0D7" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 10.300500000000001 12.555000000000001)"><use xlink:href="#g3A69B5BA14C4FC0AE0508775E710B241" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g></g><defs id="glyph"><symbol id="g97EE8E40EE0EEE4FD2AF6896DC5BA0D7" overflow="visible"><path d="M 0 0m 11.8935 9.0720005 c 0 0 0 0.14849949 -0.17549992 0.14849949 c -0.44550037 0 -0.9180002 -0.040499687 -1.3635006 -0.040499687 c -0.45899963 0 -0.9314995 0.040499687 -1.3769999 0.040499687 c -0.080999374 0 -0.24300003 0 -0.24300003 -0.2699995 c 0 -0.14850044 0.13500023 -0.14850044 0.24300003 -0.14850044 c 0.76950073 -0.013500214 0.9180002 -0.29699993 0.9180002 -0.59399986 c 0 -0.040499687 -0.026999474 -0.24300003 -0.040499687 -0.2835002 l -1.5120001 -6.0074997 l -2.9835005 7.047001 c -0.1079998 0.24299908 -0.121500015 0.2564993 -0.43199968 0.2564993 h -1.8090003 c -0.26999998 0 -0.3915 0 -0.3915 -0.2699995 c 0 -0.14850044 0.121500015 -0.14850044 0.37800002 -0.14850044 c 0.067500114 0 0.9180002 0 0.9180002 -0.121500015 l -1.8090003 -7.2495003 c -0.13499999 -0.53999996 -0.36449993 -0.972 -1.458 -1.0125 c -0.08100003 0 -0.22950006 -0.013500005 -0.22950006 -0.26999998 c 0 -0.094500005 0.067500055 -0.14850001 0.17550004 -0.14850001 c 0.43200004 0 0.9045 0.0405 1.35 0.0405 c 0.4590001 0 0.94499993 -0.0405 1.3905001 -0.0405 c 0.067499876 0 0.24300003 0 0.24300003 0.27 c 0 0.13499999 -0.121500015 0.1485 -0.26999998 0.1485 c -0.783 0.02700001 -0.89100003 0.324 -0.89100003 0.59400004 c 0 0.094499946 0.013499975 0.16199994 0.0539999 0.31050003 l 1.7820003 7.1280003 c 0.0539999 -0.08100033 0.0539999 -0.1079998 0.121500015 -0.24300003 l 3.3614998 -7.9515 c 0.094500065 -0.22950001 0.13500023 -0.2565 0.25650024 -0.2565 c 0.14849949 0 0.14849949 0.0405 0.2159996 0.28350002 l 1.8900003 7.5195 c 0.13500023 0.54000044 0.37800026 0.9585004 1.4580002 0.9990001 c 0.067500114 0 0.22949982 0.013500214 0.22949982 0.27000046 Z "/></symbol><symbol id="g3A69B5BA14C4FC0AE0508775E710B241" overflow="visible"><path d="M 0 0m 3.0618 5.90625 c 0 0.15119982 -0.11339998 0.35909986 -0.37800002 0.35909986 c -0.25515008 0 -0.5292001 -0.24569988 -0.5292001 -0.5197501 c 0 -0.16064978 0.12284994 -0.35909986 0.37800002 -0.35909986 c 0.27405 0 0.5292001 0.2645998 0.5292001 0.5197501 Z m 0.24569988 -4.5549 c 0 0.12285006 -0.12284994 0.12285006 -0.15120006 0.12285006 c -0.1322999 0 -0.14174986 -0.05669999 -0.17954993 -0.16065001 c -0.21735 -0.756 -0.63314986 -1.14345 -1.0016999 -1.14345 c -0.18900001 0 -0.23625004 0.12285 -0.23625004 0.33074996 c 0 0.21735 0.06615007 0.3969 0.15120006 0.6048 l 0.3024 0.75600004 l 0.4630499 1.20015 c 0.028350115 0.09449983 0.05669999 0.20789981 0.05669999 0.30239987 c 0 0.44414997 -0.37800002 0.8032501 -0.8977499 0.8032501 c -0.93555 0 -1.37025 -1.2852001 -1.37025 -1.4458499 c 0 -0.12285018 0.13229999 -0.12285018 0.16064999 -0.12285018 c 0.13230002 0 0.14174998 0.047250032 0.17009997 0.15120006 c 0.2457 0.81270003 0.6615 1.1529 1.01115 1.1529 c 0.15119994 0 0.23625004 -0.07559991 0.23625004 -0.33075 c 0 -0.21735 -0.05669999 -0.3591001 -0.29295003 -0.94499993 l -0.59535 -1.52145 c -0.037799954 -0.12284994 -0.08504999 -0.23624998 -0.08504999 -0.39689994 c 0 -0.44415003 0.37800002 -0.80325 0.89775 -0.80325 c 0.94500005 0 1.3607999 1.3040999 1.3607999 1.4458499 Z "/></symbol></defs></svg></span> 个点，然后对每个像素取深度上最靠前的点 <span role="math"><svg class="typst-frame" style="overflow: visible; width: 3.587566666666667em; height: 0.683em;" viewBox="0 0 48.43215 9.2205" width="48.43215pt" height="9.2205pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:h5="http://www.w3.org/1999/xhtml"><g><g class="typst-text" transform="matrix(1 0 0 -1 0 9.2205)"><use xlink:href="#g50DA4B7BDD6A87E9D1327A0524772811" x="0" y="0" fill="#000000" fill-rule="nonzero"/><use xlink:href="#gEFB0D4954C8ADA240A829471597EA46F" x="11.2455" y="0" fill="#000000" fill-rule="nonzero"/><use xlink:href="#gE2E2D5A1CCC8ABB00A2F0ABA3493DF80" x="14.9985" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 22.5045 12.555)"><use xlink:href="#g4E22239E6353EBCF122352F4110B374" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 28.20285 9.82395)"><use xlink:href="#g56ED44F5FD755E6F0095147CEA18179" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 35.70705 9.2205)"><use xlink:href="#g30BD3B82B09CBDBBB2BC34793CC2B83E" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 41.89005 4.299750000000002)"><use xlink:href="#gE25CF5D10793F9A0D0B42E2818F6348F" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g><g class="typst-text" transform="matrix(1 0 0 -1 41.552550000000004 12.555)"><use xlink:href="#g706D9051A358F9E2DAA9FFCC388B99A8" x="0" y="0" fill="#000000" fill-rule="nonzero"/></g></g><defs id="glyph"><symbol id="g50DA4B7BDD6A87E9D1327A0524772811" overflow="visible"><path d="M 0 0m 10.9755 0 v 0.4185 c -0.70199966 0 -1.0395002 0 -1.0529995 0.40500003 v 2.5785003 c 0 1.161 0 1.5795 -0.41849995 2.0655 c -0.18900013 0.22949982 -0.6345005 0.4994998 -1.4175005 0.4994998 c -1.1339998 0 -1.7280002 -0.80999994 -1.9575 -1.323 c -0.18900013 1.1745 -1.1880002 1.323 -1.7955003 1.323 c -0.98549986 0 -1.6199999 -0.58049965 -1.9979999 -1.4175 v 1.4175 l -1.9035 -0.14849997 v -0.41849995 c 0.94500005 0 1.053 -0.094500065 1.053 -0.75600004 v -3.618 c 0 -0.6075 -0.14849997 -0.6075 -1.053 -0.6075 v -0.4185 l 1.5255 0.0405 l 1.512 -0.0405 v 0.4185 c -0.9045 0 -1.053 0 -1.053 0.6075 v 2.484 c 0 1.404 0.9584999 2.16 1.8225002 2.16 c 0.85049963 0 0.9989996 -0.7290001 0.9989996 -1.4984999 v -3.1455002 c 0 -0.6075 -0.14849997 -0.6075 -1.053 -0.6075 v -0.4185 l 1.5255003 0.0405 l 1.5120001 -0.0405 v 0.4185 c -0.9045 0 -1.053 0 -1.053 0.6075 v 2.484 c 0 1.404 0.9584999 2.16 1.8224998 2.16 c 0.8505006 0 0.9990001 -0.7290001 0.9990001 -1.4984999 v -3.1455002 c 0 -0.6075 -0.14849949 -0.6075 -1.053 -0.6075 v -0.4185 l 1.5254998 0.0405 Z "/></symbol><symbol id="gEFB0D4954C8ADA240A829471597EA46F" overflow="visible"><path d="M 0 0m 3.3345 0 v 0.4185 c -0.89100003 0 -0.94499993 0.067499995 -0.94499993 0.59400004 v 4.9545 l -1.8900001 -0.14849997 v -0.41849995 c 0.87750006 0 0.99899995 -0.08099985 0.99899995 -0.7424998 v -3.6315002 c 0 -0.6075 -0.14849997 -0.6075 -1.053 -0.6075 v -0.4185 l 1.485 0.0405 c 0.4725001 0 0.94499993 -0.026999999 1.404 -0.0405 Z m -0.74250007 8.154 c 0 0.36450005 -0.3104999 0.7154999 -0.7155 0.7154999 c -0.459 0 -0.729 -0.37800026 -0.729 -0.7154999 c 0 -0.36450005 0.31050003 -0.7154999 0.7155 -0.7154999 c 0.459 0 0.729 0.37799978 0.729 0.7154999 Z "/></symbol><symbol id="gE2E2D5A1CCC8ABB00A2F0ABA3493DF80" overflow="visible"><path d="M 0 0m 7.2225003 0 v 0.4185 c -0.70200014 0 -1.0395002 0 -1.053 0.40500003 v 2.5785003 c 0 1.161 0 1.5795 -0.41850042 2.0655 c -0.18899965 0.22949982 -0.6345 0.4994998 -1.4175 0.4994998 c -0.98549986 0 -1.6199999 -0.58049965 -1.9979999 -1.4175 v 1.4175 l -1.9035 -0.14849997 v -0.41849995 c 0.94500005 0 1.053 -0.094500065 1.053 -0.75600004 v -3.618 c 0 -0.6075 -0.14849997 -0.6075 -1.053 -0.6075 v -0.4185 l 1.5255 0.0405 l 1.512 -0.0405 v 0.4185 c -0.9045 0 -1.053 0 -1.053 0.6075 v 2.484 c 0 1.404 0.9584999 2.16 1.8225002 2.16 c 0.85049963 0 0.9989996 -0.7290001 0.9989996 -1.4984999 v -3.1455002 c 0 -0.6075 -0.14849997 -0.6075 -1.053 -0.6075 v -0.4185 l 1.5255003 0.0405 Z "/></symbol><symbol id="g4E22239E6353EBCF122352F4110B374" overflow="visible"><path d="M 0 0m 5.4337497 6.4259996 c 0 0.04725027 -0.037799835 0.13230038 -0.15119982 0.13230038 c -0.18900013 0 -0.97335005 -0.07560015 -1.2096 -0.094500065 c -0.07559991 -0.009449959 -0.20790005 -0.018899918 -0.20790005 -0.21735 c 0 -0.1322999 0.13230014 -0.1322999 0.24569988 -0.1322999 c 0.4536004 0 0.4536004 -0.06615019 0.4536004 -0.14175034 c 0 -0.06614971 -0.018900394 -0.12284994 -0.037800312 -0.20790005 l -0.5386498 -2.1545997 c -0.19845009 0.31184983 -0.5197501 0.55754995 -0.97335005 0.55754995 c -1.2096 0 -2.4380999 -1.3324499 -2.4380999 -2.69325 c 0 -0.91665 0.61424994 -1.5687 1.4269499 -1.5687 c 0.51030016 0 0.9639001 0.29295 1.3419001 0.6615 c 0.17954993 -0.567 0.7276499 -0.6615 0.9733498 -0.6615 c 0.34019995 0 0.57645035 0.20789999 0.7465501 0.50084996 c 0.20790005 0.36854997 0.33075 0.9072 0.33075 0.94499993 c 0 0.12285006 -0.12284994 0.12285006 -0.15119982 0.12285006 c -0.13230038 0 -0.14175034 -0.037800074 -0.20790005 -0.29295003 c -0.11339998 -0.4536 -0.29295015 -1.01115 -0.68984985 -1.01115 c -0.24570036 0 -0.31185007 0.20789999 -0.31185007 0.46304998 c 0 0.17954999 0.018899918 0.26459998 0.047249794 0.38744998 l 1.31355 5.2542 c 0.009449959 0.018899918 0.037799835 0.15119982 0.037799835 0.15119982 Z m -1.5781498 -3.3452997 c 0 -0.037800074 -0.018899918 -0.10395002 -0.028349876 -0.1322999 l -0.44414997 -1.76715 c -0.06615019 -0.26459998 -0.27405 -0.4536 -0.48195004 -0.63315 c -0.085050106 -0.0756 -0.46305013 -0.378 -0.8694 -0.378 c -0.34965003 0 -0.68985 0.24569997 -0.68985 0.91665006 c 0 0.50084996 0.27405 1.54035 0.4914 1.91835 c 0.4346999 0.7559998 0.91664994 0.8977499 1.18125 0.8977499 c 0.6615 0 0.8410499 -0.71819997 0.8410499 -0.82215 Z "/></symbol><symbol id="g56ED44F5FD755E6F0095147CEA18179" overflow="visible"><path d="M 0 0m 3.4425 4.212 c 0 0.15525007 -0.11475015 0.2902503 -0.29699993 0.2902503 c -0.20925021 0 -0.41850019 -0.20249987 -0.41850019 -0.4050002 c 0 -0.16199994 0.12825012 -0.29024982 0.29700017 -0.29024982 c 0.22949982 0 0.41849995 0.21600008 0.41849995 0.40499973 Z m -0.24300003 -1.8832498 c 0 0.3847499 -0.35774994 0.6547499 -0.83025 0.6547499 c -0.81675005 0 -1.3162501 -0.9112499 -1.3162501 -1.04625 c 0 -0.094499946 0.10800004 -0.094499946 0.1417501 -0.094499946 c 0.10124993 0 0.11474991 0.02699995 0.14849997 0.094499946 c 0.31050003 0.6615001 0.7695 0.81675017 0.9990001 0.81675017 c 0.26999998 0 0.28349996 -0.24300003 0.28349996 -0.35775018 c 0 -0.094500065 -0.006750107 -0.121500015 -0.020250082 -0.17549992 l -0.6344999 -2.5245001 c -0.12825 -0.513 -0.55350006 -0.8505 -0.96525 -0.8505 c -0.11475003 0 -0.18900001 0.020249963 -0.24300003 0.033749938 c 0.162 0.094500065 0.20924997 0.27000004 0.20924997 0.35775006 c 0 0.11475003 -0.08099997 0.27675003 -0.30374998 0.27675003 c -0.22950003 0 -0.4185 -0.20925003 -0.4185 -0.43200004 c 0 -0.29025 0.30374998 -0.46575004 0.76275 -0.46575004 c 0.459 0 1.3364999 0.23625004 1.5457499 1.07325 l 0.6075001 2.4164999 c 0.013499975 0.067500114 0.033750057 0.12825012 0.033750057 0.22275019 Z "/></symbol><symbol id="g30BD3B82B09CBDBBB2BC34793CC2B83E" overflow="visible"><path d="M 0 0m 5.8050003 1.4445001 c 0 0.08099997 -0.08100033 0.17549992 -0.16200018 0.17549992 c -0.067500114 0 -0.094500065 -0.02699995 -0.17549992 -0.13499999 c -1.0665002 -1.3365 -2.538 -1.3365 -2.7 -1.3365 c -0.8505001 0 -1.2150002 0.6615 -1.2150002 1.4715 c 0 0.55350006 0.26999998 1.863 0.7290001 2.7000003 c 0.41849995 0.7694998 1.161 1.3499999 1.9034998 1.3499999 c 0.4590001 0 0.9720001 -0.17549992 1.1610003 -0.53999996 c -0.21600008 0 -0.4050002 0 -0.59399986 -0.18900013 c -0.21600008 -0.20249987 -0.24300003 -0.43199968 -0.24300003 -0.52649975 c 0 -0.32400036 0.24300003 -0.4725001 0.4994998 -0.4725001 c 0.3915 0 0.75600004 0.32400012 0.75600004 0.8640001 c 0 0.6615 -0.6345 1.1609998 -1.5929999 1.1609998 c -1.8225002 0 -3.6180003 -1.9305 -3.6180003 -3.8339999 c 0 -1.2150002 0.78300005 -2.2815 2.187 -2.2815 c 1.9305 0 3.0645003 1.431 3.0645003 1.593 Z "/></symbol><symbol id="gE25CF5D10793F9A0D0B42E2818F6348F" overflow="visible"><path d="M 0 0m 4.14855 5.90625 c 0 0.16064978 -0.12284994 0.35909986 -0.37800002 0.35909986 c -0.27405 0 -0.5292001 -0.2645998 -0.5292001 -0.5197501 c 0 -0.15119982 0.11339998 -0.35909986 0.37800002 -0.35909986 c 0.25514984 0 0.5292001 0.24569988 0.5292001 0.5197501 Z m -0.34965014 -2.6271 c 0 0.5197499 -0.44414997 0.88829994 -1.0111499 0.88829994 c -1.0489501 0 -1.6537501 -1.31355 -1.6537501 -1.4458499 c 0 -0.12285018 0.13230002 -0.12285018 0.16065001 -0.12285018 c 0.11339998 0 0.12285006 0.028350115 0.18900001 0.17009997 c 0.23625004 0.5481 0.7276499 1.1340001 1.2757499 1.1340001 c 0.23624992 0 0.32130003 -0.16065001 0.32130003 -0.4630499 c 0 -0.10395002 -0.018899918 -0.23625016 -0.028349876 -0.2835002 l -0.9072001 -3.6193497 c -0.16065001 -0.65205 -0.6615 -1.2001499 -1.2190499 -1.2001499 c -0.12285 0 -0.23624998 0.018899918 -0.34019995 0.05669999 c 0.24569994 0.11339998 0.32129997 0.34019995 0.32129997 0.48194993 c 0 0.22680002 -0.17954999 0.34965003 -0.36855 0.34965003 c -0.29294997 0 -0.53865 -0.25514996 -0.53865 -0.55754995 c 0 -0.3591001 0.36855 -0.59535 0.93555 -0.59535 c 0.56700003 0 1.6726499 0.34019995 1.9656 1.4930999 l 0.85049987 3.3831 c 0.028350115 0.10395002 0.047250032 0.18899989 0.047250032 0.33075 Z "/></symbol><symbol id="g706D9051A358F9E2DAA9FFCC388B99A8" overflow="visible"><path d="M 0 0m 5.50935 3.4775999 c 0 0.52919984 -0.60479975 0.6898501 -0.9828 0.6898501 c -0.47249985 0 -0.85049987 -0.31185007 -1.0583999 -0.67094994 c -0.17009997 0.38744998 -0.6048 0.67094994 -1.1340001 0.67094994 c -1.0583999 0 -1.6631999 -1.1623499 -1.6631999 -1.4458499 c 0 -0.12285018 0.13229996 -0.12285018 0.16064996 -0.12285018 c 0.12285 0 0.13230002 0.037800074 0.17955005 0.16065001 c 0.23624992 0.7371001 0.83159995 1.14345 1.29465 1.14345 c 0.31185007 0 0.5669999 -0.17009997 0.5669999 -0.6236999 c 0 -0.18900013 -0.11339998 -0.6615 -0.19844985 -0.9828 l -0.32130003 -1.2946501 c -0.094500065 -0.3212999 -0.40635002 -0.83159995 -0.88830006 -0.83159995 c -0.028349996 0 -0.31184995 0 -0.5103 0.13229999 c 0.38745004 0.12285 0.41580003 0.46305 0.41580003 0.51975 c 0 0.20789993 -0.16065001 0.34965003 -0.37800002 0.34965003 c -0.26459998 0 -0.53865 -0.22680002 -0.53865 -0.57645005 c 0 -0.47249997 0.5292 -0.68985 0.99225 -0.68985 c 0.4346999 0 0.8221499 0.27405 1.0584 0.67095 c 0.22679996 -0.50084996 0.74654984 -0.67095 1.1245499 -0.67095 c 1.08675 0 1.6631999 1.18125 1.6631999 1.4458499 c 0 0.12285006 -0.12284994 0.12285006 -0.15119982 0.12285006 c -0.1322999 0 -0.14175034 -0.047250032 -0.17955017 -0.16065001 c -0.19845009 -0.65205 -0.76545 -1.14345 -1.2946498 -1.14345 c -0.37800002 0 -0.5764501 0.25515 -0.5764501 0.61424994 c 0 0.24569994 0.22679996 1.10565 0.4914 2.1546 c 0.18900013 0.72765017 0.61424994 0.9639001 0.9261 0.9639001 c 0 0 0.31185007 0 0.51030016 -0.1322999 c -0.3024001 -0.085050106 -0.4158001 -0.34965014 -0.4158001 -0.5197501 c 0 -0.20790005 0.16064978 -0.3496499 0.37799978 -0.3496499 c 0.21735 0 0.5292001 0.17954993 0.5292001 0.5764499 Z "/></symbol></defs></svg></span> 作为 Splat 结果的无偏估计值。</p>