ADM-201 dump PMP dumps pdf SSCP exam materials CBAP exam sample questions

AI改进烟雾模拟 – 译学馆
未登录,请登录后再发表信息
最新评论 (0)
播放视频

AI改进烟雾模拟

AI Learns To Improve Smoke Simulations | Two Minute Papers #188

亲爱的学霸们 这是由Károly Zsolnai-Fehér带来的两分钟论文
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
这期我们要介绍的是 利用AI创建超级精细的烟雾模拟
This work is about using AI to create super detailed smoke simulations.
一般来说 初步模拟用不了很长时间 但随着分辨率的提高
Typically, creating a crude simulation doesn’t take very long, but as we increase the resolution,
执行时间和内存消耗也在飙升
the execution time and memory consumption skyrockets.
在人工智能的时代 似乎只有向过程中添加一些学习算法 才能合理地解决
In the age of AI, it only sounds logical to try to include some learning algorithms in
这个问题
this process.
假如有一项基于AI的技术懂得一点烟雾模拟 并且可以接受原始数据
So what if we had an AI-based technique that would have some sort of understanding of smoke
然后向其中添加细节 那会怎样呢?
simulations, take our crude data and add the fine details to it?
按照这种方法 不需要几天或是几星期的计算 我们就可以得到
This way, we could obtain a high resolution smoke simulation without waiting several days
高分辨率的烟雾模拟
or weeks for the computation.
现在 如果你真的是一个经验丰富的学霸 你可能会记得一篇叫做Wavelet Turbulence的早期作品
Now if you are a truly seasoned Fellow Scholar, you may remember an earlier work by the name
这是我有史以来最喜欢的一篇论文
Wavelet Turbulence, which is one of my favorite papers of all time.
以至于它在第一期两分钟论文中
So much so that it got the distinction of being showcased in the very first Two Minute
被展示出来
Papers episode.
当我第一次看到它时 我还是一名大二的学生
I was a sophomore college student back then when I’ve first seen it and was absolutely
当时就被它最终的效果震撼到了
shocked by the quality of the results.
那简直令我难以忘却
That was an experience I’ll never forget.
它也因此获得了奥斯卡技术奖 不夸张地说 这是有史以来最具有影响力的作品之一
It also won a technical Oscar award and it is not an overstatement to say that this was
这件事使我意识到 研究才是我兴趣所在
one of the most influential works that made me realize that research is my true calling.
如果你想看看它有多么震撼 可以翻出来看看
The link to the first episode is available in the video description and if you want to
第一集的链接在视频说明中就有喔!
see how embarrassing it is, make sure to check it out.
它并没有使用AI来完成类似的事情 而是使用启发法
It did something similar, but instead of using AI, it used some heuristics that describe
来描绘出一个流体或烟雾中较小涡流和较大涡流的比例和分布
what is the ratio and distribution of smaller and bigger vortices in a piece of fluid or smoke.
应用这些信息 它可以创建出一些相似的效果 但基本上
Using this information, it could create a somewhat similar effect, but ultimately, that
这种技术只了解一般的烟雾模拟 而对于我们手头上某个具体的
technique had an understanding of smoke simulations in general, but it didn’t know anything about
烟雾场景却一无所知
the scene that we have at hand right now.
与此相关的另一项工作 是向AI展示一段烟雾模拟视频
Another work that is related to this is showing a bunch of smoke simulation videos to an AI
并教会它继续进行这些模拟
and teach it to continue these simulations by itself.
我们将这项工作作为一个中间解决方案 因为这项工作表示我们应该退后一步
I would place this work as a middle ground solution, because this work says that we should
而不是从头开始合成一切
take a step back and not try to synthesize everything from scratch.
我们创建了一个模拟数据库 将它们分成一个个很小的碎片
Let’s create a database of simulations, dice them up into tiny tiny patches, look at the
分别在低分辨率和高分辨率下看同一个视频 从而学习它们之间是如何相互关联的
same footage in low and high resolutions, and learn how they relate to each other.
这样 我们就可以将一些低分辨率的镜头交给神经网络
This way, we can hand the neural network some low resolution footage and it will be able
然后它就可以有根据地推测出 哪一个高分辨率碎片最为匹配
to make an educated guess as to which high resolution patch should be the best match for it.
当我们找到正确的碎片时 只需将粗糙模拟切换到
When we found the right patch, we just switch the coarse simulation to the most fitting
数据库中最合适的高分辨率碎片下即可
high-resolution patch in the database.
你可能会说 在理论上 创造这样一个Frankenstein烟雾模拟
You might say that in theory, creating such a Frankenstein smoke simulation sounds like
听上去似乎不靠谱
a dreadful idea.
但是看看这结果 绝对很精彩!
But have a look at the results, as they are absolutely brilliant!
正如你看到的 它在一个很粗糙的基础模拟上 添加了很多细节
And as you can see, it takes a really crude base simulation and adds so many details to
这真是一项令人叹服的成就
it, it’s truly an incredible achievement.
一个神经网络被训练用于捕获密度的相似性 而另一个用于涡度
One neural network is trained to capture similarities in densities, and one for vorticity.
依次使用这两个神经网络 我们能在低分辨率流体流动的基础上
Using the two neural networks in tandem, we can take a low resolution fluid flow and synthesize
合成出微小的细节 真是让人难以置信的方法!
the fine details on top of it in a way that is hardly believable.
它也可以处理边界情况 那意味着即使我们的烟雾团碰到了物体
It also handles boundary conditions, which means that these details are correctly added
这些细节也能被正确添加进来
even if our smoke puff hits an object.
这(边界问题)对于小波湍流是个问题 必须通过几个后续工作来解决
This was an issue with Wavelet Turbulence which had to be addressed with several followup works.
这里还有对这种传奇算法的比较 你能看到
There are also comparisons against this legendary algorithm, and as you can see, the new technique
新技术要比它更好
smokes it.
然而 它足足花了9年时间才完成
However, it took 9 years to do this.
这正是世界上九大永恒研究之一
This is exactly 9 eternities in the world of research, which is a huge testament to
它完美的证实了最初的算法是多么强大
how powerful the original algorithm was.
从中我获取到越来越多的知识 也让我更加了解到学霸们的世界
It is also really cool to get more and more messages where I get to know more about you
这真的很酷!
Fellow Scholars.
我还了解到 这个系列被应用到巴西的一些学校课程中 也被用于加强大学教育
I was informed that the series is used in school classes in Brazil, it is also used
还被当作晚餐时的有趣的家庭对话主题
to augment college education, and it is a great topic for fun family conversations over dinner.
这简直太棒了
That’s just absolutely fantastic.
很高兴能得知 这个系列是你们很多人的灵感来源
Loving the fact that the series is an inspiration for many of you.
感谢您的收看和大力支持 我们下期再见!
Thanks for watching and for your generous support, and I’ll see you next time!

发表评论

译制信息
视频概述

利用AI来学习高低分辨率的模拟镜头是如何匹配的,从而对烟雾模拟进行细节化处理。

听录译者

收集自网络

翻译译者

[B]无牙无耳

审核员

审核团O

视频来源

https://www.youtube.com/watch?v=Mu0ew2F-SSA

相关推荐