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

深度照片风格转化 – 译学馆
未登录,请登录后再发表信息
最新评论 (0)
播放视频

深度照片风格转化

Deep Photo Style Transfer | Two Minute Papers

亲爱的学霸们 大家好 这里是由Károly Zsolnai-Fehér带来的两分钟论文
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
让我们来看看这个关于照片风格转化的宏伟技术
Let’s have a look at this majestic technique that is about style transfer for photos.
风格转化是一种神奇的算法 它需要我们有一张有内容的照片和一张
Style transfer is a magical algorithm where we have one photograph with content, and one
风格有趣的照片
with an interesting style.
而输出的是这两张照片融合而成的第三张照片
And the output is a third image with these two photos fused together.
这通常是用一种经典的名为卷积神经网络的机器学习技术
This is typically achieved by a classical machine learning technique that we call a
完成的
convolutional neural network.
这些网络包含的层数越多 功能就越强
The more layers these networks contain, the more powerful they are, and the more capable
并且它们建立对图像的直观理解的能力也越强
they are in building an intuitive understanding of an image.
我们早期的几次节目是关于可视化神经网络的内部运作的
We had several earlier episodes on visualizing the inner workings of these neural networks,
同往常一样 可以在视频描述中找到相关链接
as always, the links are available in the video description.
不要错过 我相信你会像我第一次看到结果时
Don’t miss out, I am sure you’ll be as amazed by the results as I was when I have first
那么惊讶
seen them.
如果我们正在寻求的是一个绘画风格的结果 那之前的这些神经网络风格转化技术
These previous neural style transfer techniques work amazingly well if we’re looking for a
就做得非常好
painterly result.
然而 对于照片风格转化 这里的特写镜头显示它们引入了不必要的
However, for photo style transfer, the closeups here reveal that they introduce unnecessary
扭曲图像
distortions to the image.
它们看起来一点也不写实了
They won’t look realistic anymore.
但这新的一个不是
But not with this new one.
看看这些结果
Have a look at these results.
这是绝对疯狂的
This is absolute insanity.
它们在某种意义上是正确的
They are just right in some sense.
它们有一个难以捉摸的特点
There is an elusive quality to them.
还有 这个简直就是挑战!
And this is the challenge!
我们不仅必须将我们要寻找的东西放在一边 而且我们必须找到一个
We not only have to put what we’re searching for into words, but we have to find a mathematical
一个计算机能够执行的数学描述
description of these words to make the computer execute it.
那么这个定义是什么呢?
So what would this definition be?
只要想想 这是一个非常具有挑战性的问题
Just think about this, this is a really challenging question.
作者决定 要将输出图像的真实感最大化
The authors decided that the photorealism of the output image is to be maximized.
好 这听起来不错 但谁又确实知道真实感的严谨的数学描述是什么呢?
Well, this sounds great, but who really knows a rigorous mathematical description of photorealism?
一个可能的解决办法是 规定输出颜色的改变
One possible solution would be to stipulate that the changes in the output color would
必须保留输入样式颜色的比例和距离
have to preserve the ratios and distances of the input style colors.
类似的规则用于线性代数和计算机图形
Similar rules are used in linear algebra and computer graphics to make sure shapes don’t
以确保当我们对它们进行旋转、转换和其它更多操作时 图形不会被扭曲
get distorted as we’re tormenting them with rotations, translations and more.
我们喜欢称这些操作为仿射变换
We like to call these operations affine transformations.
所以完全科学的描述就是 我们在公式中添加一个正则项
So the fully scientific description would be that we add a regularization term that
规定这些颜色只能进行仿射变换
stipulates, that these colors only undergo affine transformations.
但是 我们在这里又说了一个新词 这个正则项是什么意思?
But we’ve used one more new word here – what does this regularization term mean?
它意味着转换颜色有许多不同可能的解决方案
This means that there are a ton of different possible solutions for transferring the colors,
而我们试图引导优化算法采取那些符合一些额外标准的解决方案
and we’re trying to steer the optimizer towards solutions that adhere to some additional criterion,
在我们的例子中 额外标准就是指仿射变换
in our case, the affine transformations.
在这个问题的数学描述中 这些额外的规定以
In the mathematical description of this problem, these additional stipulations appear in the
正则项的形式出现
form of a regularization term.
我很高兴你们这些同学一直在看两分钟论文
I am so happy that you Fellow Scholars have been watching Two Minute Papers for so long,
所以我们终于可以谈论这样的技术了
that we can finally talk about techniques like this.
让观众对这些话题有一定的了解是非常棒的
It’s fantastic to have an audience that has this level of understanding of these topics.
喜欢它
Love it.
只是绝对地喜欢它
Just absolutely love it.
该项目的源代码同样也是可用的
The source code of this project is also available.
另外 一定要看看Distill 一个绝对让人惊讶的新科学期刊
Also, make sure to have a look at Distill, an absolutely amazing new science journal
它来自谷歌大脑团队
from the Google Brain team.
不过这不是普通的期刊 因为他们正在寻找的并不是新技术
But this is no ordinary journal, because what they are looking for is not necessarily novel
而是以新颖直观的方式解释现有的技术
techniques, but novel and intuitive ways of explaining already existing works.
它们有一个很好的“研究债务”专栏 几乎可以被理解成
There is also an excellent write-up on research debt that can almost be understood as a manifesto
是这个期刊的宣言
for this journal.
一个确实值得一读的期刊
A worthy read indeed.
他们还设立了一个“科学蒸馏”奖项
They also created a prize for science distillation.
我喜欢这个新举措 相信在不久的将来
I love this new initiative and I am sure we’ll hear about
就会听到很多关于这个期刊的消息
this journal a lot in the near future.
确保你会看一看 在视频描述中有所有这些内容的链接
Make sure to have a look, there is a link to all of these in the video description.
感谢您的观看和慷慨支持 下次再见!
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

本次两分钟论文的主题是一种照片风格转换的宏伟技术

听录译者

收集自网络

翻译译者

B11101001

审核员

知易行难

视频来源

https://www.youtube.com/watch?v=HTUxsrO-P_8

相关推荐