3D transparency proposal

未关闭
  • 哪里不对？
#8,112 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 2 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
8/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
javascript
领域
data-visualization

调研方向

This is a design proposal for screen-space blue-noise transparency in plotly.js's 3D surfaces, and the payload names no files, tests or entry points. Start by reading the linked issues (#1267, #3243, #4918) to see what has already been tried. Done would mean a maintainer-agreed design and a prototype, so a first-time contributor should discuss the approach in the thread before writing code.

由索引模型根据 Issue 内容生成。

描述

enhancement feature P3 plotly-internal size: 5

3D transparency has always been tricky, and has never worked very well in plotly.js. - see eg https://github.com/plotly/plotly.js/issues/1267 https://github.com/plotly/plotly.js/issues/3243 https://github.com/plotly/plotly.js/issues/4918 for just a few examples. Several years ago I played with an approach that seemed promising - treating opacity as a semi-random probability of drawing or not drawing a pixel, then oversampling / averaging to smooth the result. I never got it working well enough to include in the library, but in another project (not using plotly.js) a colleague of mine has refined the concept into what I think is a really beautiful, performant, and useful result so I'm sharing here in case someone wants to bring it into plotly.js

  • Each transparent surface gets a repeatable, screen-space random blue noise function that yields values between 0 and 1. This means if you pass through multiple layers of the same surface we will not make it as opaque as it really should be (because they all use the same screen-space function), but because of the next point, the result still looks physically reasonable, and arguably more useful than the full opacity of a complete physical model. If you don't use screen space for the noise function (use full 3D space for example) then there's a distracting amount of motion to the noise as you move the viewport.
  • The transparency at each pixel becomes T^(1/cos(theta)) where T is the given transparency and theta is the angle at that pixel between the surface normal and the camera. This makes shallow-incidence areas more opaque, modeling a physical system where you're passing through a greater thickness of the material depending on this angle.
  • Only pixels where the final opacity is greater than the blue noise function (and where we haven't already drawn a surface closer to the camera) are drawn. All the same lighting effects used for opaque surfaces can still be used.
  • During mouse interactions this is what we display, so it looks grainy but you still see the features and on average they are correct. This is one area I still think the algorithm we're using could be improved, it often looks kind of like those old "Magic Eye" books whereas I think with some experimentation we could get it a bit flatter, but it's entirely usable in this form.
  • Once the mouse stops moving we average many frames (in the video below we use 64 frames) of this with different random seeds on the blue noise functions, to smooth away the graininess. In the averaging process it's important to distinguish between sRGB and linear color.

https://github.com/user-attachments/assets/29266885-2745-4077-a7db-33d24d96daee

主要语言
JavaScript
星标
18.4k
派生
2.1k
平均合并
1 天 5 小时
30 天内合并 PR
12

环境准备

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

plotly/plotly.js 的其他 Issue

查看 plotly/plotly.js 的全部 Issue

相似的 Issue

更多 JavaScript Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。