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

AI已经学会合成动物图片 – 译学馆
未登陆,请登陆后再发表信息
最新评论 (0)
播放视频

AI已经学会合成动物图片

AI Learns to Synthesize Pictures of Animals | 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.
我刚读完这篇论文 就惊讶得从椅子上掉了下来
I just finished reading this paper and I fell out of the chair.
我几乎可以保证 这项工作的结果是如此疯狂 你将不得不两遍
And I can almost guarantee you that the results in this work are so insane, you will have
甚至三遍的检查 以确信你现在看到的这些东西
to double, or even triple check to believe what you’re going to see here.
这一篇是关于图像转换的 它意味着输入一个图像
This one is about image translation, which means that the input is an image, and the
再根据我们的指示 输出这个输入图像的不同版本
output is a different version of this input image that is changed according to our guidelines.
设想一下 我们有一幅莫奈的画 我们想要以此来创造一幅美丽景色的照片
Imagine that we have a Monet painting, and we’d like to create a photograph of this beautiful view.
好了
There we go.
如果我们想将这个冬季景观变成一个夏天里的图像呢?
What if we’d like to change this winter landscape to an image created during the summer?
好了
There we go.
如果我们像互联网论坛上的一些人一样 喜欢比较两种完全不同的事物
If we are one of those people on the internet forums who just love to compare apples to
现在这也是有可能的
oranges, this is now also a possibility.
看看这个 设想我们喜欢这张图片的背景
And have a look at this – imagine that we like the background of this image, but instead
但比起图上的斑马 我们更想要的是几匹马
of the zebras, we would like to have a couple of horses.
没问题
No problem.
马上就好!
Coming right up!
该算法会从头合成它们
This algorithm synthesizes them from scratch.
对于这项技术 第一件我们应该了解的重要的事情 是它使用的
The first important thing we should know about this technique, is that it uses generative
生成式对抗网络
adversarial networks.
这意味着我们有两个神经网络在军备竞赛中相互竞争
This means that we have two neural networks battling each other in an arms race.
生成器网络试图创建越来越多逼真的图像
The generator network tries to create more and more realistic images, and these are passed
这些图片被传递到识别网络 从而尝试了解真正的照片
to the discriminator network which tries to learn the difference between real photographs
和虚假伪造的图像之间的差异
and fake, forged images.
在这个过程中 两个神经网络一起学习和改进
During this process, the two neural networks learn and improve together until they become
直到它们成为各自方面的专家
experts at their own craft.
然而 这件作品在这个过程中引入了两个新的内容
However, this piece of work introduces two novel additions to this process.
第一 在早期的作品中 训练样本通常是配对的
One, in earlier works, the training samples were typically paired.
这意味着一张鞋子的照片将与描绘它的图纸配对
This means that the photograph of a shoe would be paired to a drawing that depicts it.
这些附加信息非常有助于训练过程
This additional information helps the training process a great deal and the algorithm would
并且该算法将能够把图纸映射到照片
be able to map drawings to photographs.
然而 这里的一个重要区别是 没有这样的配对 我们不需要这些标签
However, a key difference here is that without such pairings, we don’t need these labels,
我们可以在我们的数据集中使用更多的训练样本 这也有助于学习进程
we can use significantly more training samples in our datasets which also helps the learning process.
如果这样执行得很好 那么该技术就可以在任何其它东西之间进行配对
If this is executed well, the technique is able to pair anything to anything else, which
这将产生一个非常强大的算法
results in a remarkably powerful algorithm.
第二个关键差异 一个循环一致性损失函数被引入到优化问题里
Key difference number two – a cycle consistency loss function is introduced to the optimization problem.
这意味着如果我们将夏季图像转换为冬季图像 然后再回到夏季图像
This means that if we convert a summer image to a winter image, and then back to a summer
我们应该获得相同的输入图像
image, we should get the very same input image back.
如果我们的学习系统遵循这一原则 转换的输出质量
If our learning system obeys to this principle, the output quality of the translation is going
将会显著提高
to be significantly better.
该循环一致性损失作为一个正则项被引入
This cycle consistency loss is introduced as a regularization term.
熟悉我们的学霸们已经知道这是什么意思 但万一你不理解
Our seasoned Fellow Scholars already know what it means, but in case you don’t, I’ve
我也已经在视频说明中给出了解释链接
put a link to our explanation in the video description.
这篇论文包含了大量其它的结果 幸运的是 该项目的源代码也是可用的
The paper contains a ton more results, and fortunately, the source code for this project is also available.
多个实现版本 真的!
Multiple implementations, in fact!
另外补充一下 令人惊讶的是 它还附带支持一些基本的视频功能
Just as a side note, which is jaw dropping, by the way – there is some rudimentary support for video.
惊人的工作 好样的!
Amazing piece of work. Bravo!
现在你也可以看到 机器学习研究的进展速度
Now you can also see that the rate of progress in machine learning research
真的是太快了!
is completely out of this world!
毫无疑问 这将是成为研究科学家的最佳时机
No doubt that it is the best time to be a research scientist it’s ever been.
如果您喜欢这一集 请务必订阅该系列
If you’ve liked this episode, make sure to subscribe to the series and have a look at
并查看我们的Patreon页面 您可以在其中找到很酷的福利 比如 在早期链接里
our Patreon page, where you can pick up cool perks, like watching every single one of these
查看每一个剧集
episodes in early access.
感谢您的观看和慷慨支持 我们下次再见!
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

用生成对抗网络(GAN)来进行图像转换,AI已经学会合成动物图片。

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=D4C1dB9UheQ

相关推荐