完美构图：Topaz Labs 首席技术官杨阿尔伯特将提升您的照片品质 - 第126期

你好，欢迎收听NVIDIA人工智能播客。

Hello, and welcome to the NVIDIA AI podcast.

我是您的主持人诺亚·克拉维茨。

I'm your host, Noah Kravitz.

过去，当我们在电视或电影院观看科幻惊悚片时，这曾是居家科技爱好者们最爱的消遣。

It used to be a favorite pastime of armchair tech experts while we watch sci fi thrillers on TV or at the movies.

故事中的英雄们会在实验室里，查看监控录像或审视监视行动拍摄的照片。

The heroes of the story would be in the lab, reviewing surveillance footage or looking at photos from a stakeout.

其中一人会发现一些有趣的东西，也许正是能彻底揭开整个案件真相的关键线索，但那画面太小或太模糊，无法清晰辨认。

One of them would spot something interesting, maybe just the thing that would blow the whole case wide open, but it'd be too small or too blurry to make out clearly.

于是我们的英雄会双击屏幕，或者只是在空中挥动手臂并下达指令。

So our hero would double tap the screen or maybe just wave her hands in the air while commanding.

放大，增强，画面便会神奇地扩展并变得清晰，同时解开谜团。

Zoom in, enhance, And the image would magically expand and come into focus, solving the mystery at the same time.

这时，像我这样的居家专家就会对着电视大喊：这不可能！

At which point, the armchair experts like me would yell at the TV screen, That's impossible!

这种技术根本做不到！

The tech can't do that!

但现在，也许它真的可以了。

Except now, maybe it can.

我今天的嘉宾是杨博士。

My guest today is Doctor.

冯·阿尔伯特·杨，Topaz Labs的创始人兼总裁。

Feng Albert Yang, founder and president of Topaz Labs.

Topaz Labs在全球专业摄影师中享有盛誉，他们正在开创深度学习与照片降噪技术的融合应用。

Topaz Labs is a well known name among serious photographers around the world, and they're pioneering the intersection of deep learning and photo noise reduction.

Topaz提供了一整套基于人工智能的工具，能够为静态图像和视频实现各种神奇的处理效果。

Topaz has a full suite of AI powered tools for working all kinds of magic on still images and video footage.

阿尔伯特将与我们探讨深度学习和GPU如何改变摄影世界。

And Albert is here to talk about how deep learning and GPUs are changing the world of photography.

他还会告诉我们，为什么如果你在德克萨斯州创业，可能会选择在客房而不是车库里起步。

He's also going to tell us why if you're starting a company in Texas, you might do it in your guest room and not in the garage.

我们稍后再谈这个。

We'll get to that later.

阿尔伯特，非常感谢你抽出时间参加NVIDIA AI播客。

Albert, thanks so much for making the time to join the NVIDIA AI podcast.

欢迎。

Welcome.

非常感谢。

Thank you very much.

这真是一个很好的介绍。

That is a really nice introduction.

无论我们是否是摄影师，都能感同身受电影里那种场景——你轻点屏幕，就能清晰地看到发生了什么。

Well, it's a it's it's it's something we can all relate to whether we're photographers or not, that that image from the movies where, you know, you tap the screen and magically just just see clearly what's going on.

但也许，多亏了你们的工作，这现在真的有可能实现了。

But actually, maybe maybe it's possible now, thanks to the work that you're doing.

所以给我们讲讲Topaz Labs是做什么的，你们具体做些什么。

So tell us a little bit about what Topaz Labs is and and what you do.

Topaz Labs 是一家专注于为摄影师和视频创作者开发 PC 和 Mac 工具的小公司。

Topaz Labs is a company is a small company focusing on making PC and Mac tools for photographer and videographer.

这家公司成立于十五年前。

The company was founded fifteen years ago.

正如 Brian 精彩地描述的那样，我最初在德克萨斯州的卧室里把它当作一个爱好开始的。

As Brian nicely put it from my spare bedroom in Texas as a hobby.

从那以后，公司发展得非常快。

Since then the company has grown a lot.

我们的优势在于，将最新的技术成果转化为摄影师能够实际使用的工具。

We basically our strength is focusing on using the most recent technology development and turning it into a usable tool for photographers.

最近，正如大家所知，人工智能，尤其是在这一领域的深度学习方面，取得了巨大进展。

Recently, as everybody know, there's a huge development in AI, especially in deep learning in this field.

是的。

Right.

图像领域出现了很多令人惊叹的应用。

And there are so many cool stuff in image area.

所以我们的公司开始专注于这个领域。

So our company start to focus on this area.

我们是第一个将基于深度学习的超分辨率技术，以及利用深度学习进行照片降噪和去模糊的产品化推向市场的公司。

And we are the first one to come up with the productization of super resolution using deep learning and also using deep learning for noise reduction and blurry removal for photo.

这变得非常流行，Topaz Labs 在这个领域获得了大量关注。

That has become sort of become so popular Topaz Labs start to get a lot of publicity in this area.

你提到了超分辨率。

You mentioned super resolution.

什么是超分辨率？深度学习是如何使其成为可能的？

What is that and and how does deep learning make it possible?

超分辨率一般来说是指放大图像或照片。

Super resolution in general means you basically upscale image or photo.

以前在研究领域，超分辨率有特定的含义。

A while ago in research area, super resolution have a special meaning.

通常是指利用多张图像中的更多信息，综合这些图像的数据，从而在更大的图像中恢复细节。

Basically, it's usually means you use more information from multiple images so that you can synthesize the information from multiple images and then recover the detail in a larger image.

好的。

Okay.

这项技术最早是在冷战时期开发的，我想。

The technology was first developed early on, I think during the Cold War.

那时，这个术语可能就开始使用了。

That was at that time the term probably got started.

首先，理论上，它永远不能用于图像增强，因为你只有一张图片。

And first, it can never, in theory, it can never use it for image enhancement because you only have one picture.

如果你将一张图片放大400%，那么对于图像中的每一个像素，你都需要生成周围额外的15个像素来实现400%的放大。

If you upsample one picture from say 400%, you basically for every single pixel in the image, you need to generate a fixed 15 more pixels around it to achieve 400% zoom up.

好的。

Okay.

理论上，这是不可能的。

And theoretically, it's just impossible.

对。

Right.

对。

Right.

而人工智能彻底改变了这一解决方案，改变了整个方程式。

And AI totally changed the solution, changed the equation.

那么去噪和去模糊是否也有类似的效果？

And so is there a similar effect with denoising and de blurring?

请原谅，我不是摄影师，我确信‘去模糊’这个说法不是最准确的术语。

And forgive me, I'm not a photographer, so I'm sure there's a better term than de blurring.

但它的效果是类似的吗？

But is it a similar kind of effect?

是的。

It is.

这类图像增强通常属于所谓的逆问题。

Those type of image enhancement is generally related, so called the inverse problem.

从一张清晰、优质的图片变成一张糟糕的图片是非常容易的。

It is very easy to make a picture from a clear picture, good picture, into a bad one.

例如，高分辨率变成低分辨率，清晰的图片变成噪点图片。

For example, high resolution become a low resolution or clear picture become a noisy picture.

你基本上是生成了一些噪声。

You basically have some generate noise.

但反过来，从噪点图片恢复成清晰图片是一个非常困难的问题。

But otherwise, from a noisy picture into a clear picture is a very hard problem.

在不丢失所有细节的前提下，你能抑制多少噪声，这存在一个理论极限。

And there is a theoretical limit on how much noise you can suppress without removing all the details.

传统上，我们都使用各种手工设计的算法来处理这类问题。

And traditionally we all use different handcraft algorithm to do those type of things.

实际上，我们在这些领域已经接近了理论极限。

And actually we were hitting the theoretical limit in those areas.

几年前，深度学习出现了，再次改变了整个局面。

And then a few years ago, Brahma comes through the deep learning and again it changed the whole equation.

那你最初是怎么开始的？

And so how did you get started?

你是怎么接触到深度学习的？又是怎么开始的？我不是要你透露机密，但你是如何开始用它来实现你对图像的那些目标，并找到突破的？

How did you find out about deep learning and start and I'm not asking you to divulge secrets, but how did you start to work with it and kind of find a breakthrough to do the things you wanted to do with images?

有几个因素起到了重要作用。

There are a few factor play an important role.

第一是我们公司的人员和文化，它们以责任感和主动精神为核心。

Number one is our company's people and the culture, which is centered around ownership and taking initiative.

我们公司非常鼓励人们去实验，我们通常会招聘一些博士生，他们只是阅读各种学术论文，寻找一些有潜力的方向。

Our company is sort of really encouraging people to experiment And we tend to hire a few PhD and they just read around academic papers and find something cruel.

所以几年前，整个AI领域其实只是有人发现了一篇论文，当时图像风格迁移非常流行。

So the whole AI thing actually a few years ago, there's just a who find a paper, at that time it was very popular to do image style transfer.

你给一张自己家的照片，再给一张比如梵高的《星夜》照片。

You give a picture of your house and then give a picture of, for example, starry night from Van Gogh.

是的。

Yeah.

它是一张类似梵高风格的图片。

It's a picture like a like a Van Gogh one.

对吧？

Right?

没错。

Right.

那是一个非常流行的项目。

That was a very popular one.

这个家伙基本上在业余时间做了个原型，大家都说：哇。

And this guy basically take this one and just in his spare time and do a prototype and everybody will say, wow.

然后我们开始将其产品化。

And we start to productize it.

没错。

Right.

同样的事情也发生在超像素上采样和降噪上。

The same thing happened to the gigapixel, which is up sampling and noise reduction.

我们做的就是大量阅读论文。

What we do is we basically read a lot of papers.

是的。

Mhmm.

每当我们看到一些有趣的论文和出色的结果时，我们就会分析它们，综合不同论文和想法，然后将它们转化为产品。

And whenever we see there are some interesting paper and a really good result, we will analyze it, combine different result, different papers, idea, and then turn them into product.

我现在正在看Topaz Labs的网站，topazlabs.com，你们有一系列插件。

And so I'm looking at the Topaz Labs website right now, topazlabs.com, and you've got a bunch of plug ins.

你们有Mask AI、Adjust AI、Denoise AI、Sharpen、Gigapixel，还有Video Enhance AI。

You've got Mask AI, Adjust AI, Denoise AI, Sharpen, Gigapixel, and then you've also got Video Enhance AI.

所以我在想，当你们从处理静态图像转向处理视频素材时，会增加多少复杂性？你们是使用了完全相同的技术，还是说处理视频完全是另一回事？

And so I'm wondering when when you switch from working with a still image to working with video footage, How much more complexity does that introduce or did you have to you know, are you using fundamentally the same techniques or is it a whole different ballgame working with video?

从某种意义上说，技术是相似的。

In one sense, it is a similar technology.

我们全都使用深度神经网络，基本思路也差不多。

We all use a very deep neural network and use pretty much a similar basic idea.

不过，我们在视频领域还比较新。

However, we are pretty new in the video area.

好的。

Okay.

一个一年前刚发布的产品。

A product that just released in a year.

我们发现视频实际上有自己的巨大挑战。

And we find video actually have its own big challenge.

是的。

Yeah.

对于视频来说，任何瑕疵因为画面在移动都会非常明显。

For video, any artifact, because it's moving, is very obvious.

在静态图像中，我们可以容忍一定程度的瑕疵，因为人们通常注意不到。

In image, we can get away with certain amount of artifact that we generally people just don't notice.

根本看不到。

Just don't see it.

对吧？

Right?

是的。

Yeah.

在视频中，任何微小的瑕疵，尤其是闪烁，人们都会立刻注意到。

In video, any small artifact, especially flickering, people immediately notice.

因此，我们在这个领域投入了大量精力。

And so we are actually working very hard in this area.

对。

Yeah.

你们最流行的插件是哪个？

What what's your most popular of the plugins?

我们目前的插件都与增强功能相关。

Our current plugin is all enhancement related.

Gigapixel 去噪和 Sharpen AI 是最受欢迎的。

That Gigapixel denoise and Sharpen AI, the most popular one.

主要是专业摄影师和那些将图像作为工作一部分的人在使用吗？

And is it predominantly professional photographers and people, you know, who do who use imagery as part of their work?

还是更多是爱好者？

Or is it more hobbyist?

你们的用户是混合的吗？

You get a mix?

我们的用户群体构成非常均衡。

We got a very healthy mix.

是的。

Yeah.

我认为大多数是专业人士和热情的用户。

The I would think majority are professional and enthusiastic.

我说过，我们的用户群体主要是专业人士，摄影是他们的专业领域。

I would say our user base is majority is professional and photography is field of ethics.

你提到公司是十五年前成立的，但我明白深度学习技术是最近才应用的。

So you mentioned that the company was founded, I think you said fifteen years ago, but it was much more recently obviously that that deep learning came into play.

你能谈谈GPU在你们的工作中所起的作用吗？

Can you talk a little bit about the role of, GPUs in what you do?

你知道，节目中很多嘉宾都会谈到，在过去五年，尤其是最近几年，深度学习技术以及利用GPU提升AI速度和准确性的进展简直呈爆炸式增长。

And, you know, lot of the guests on the show will talk about how in the past, you know, five years or and and then in particular, even the past couple of years, the the technology, the rate of acceleration and the advancements of deep learning and using GPU power to to, you know, make AI faster and more accurate has really just exploded.

你有没有发现类似的影响？

Have you found that kind of a similar effect?

或者说，最近这几年对Topaz来说具体是怎样的？

Or, you know, what have those past couple of years in particular been like for Topaz?

嗯哼。

Uh-huh.

这是我最喜爱的话题。

This is my my favorite topic.

我总是对游戏级GPU能提供的计算能力感到惊叹，更不用说专门用于机器学习的GPU了。

I'm always I'm never stopped to be amazed by how much processing power just a game game grade GPU can provide, not alone the the machine learning specific GPU.

我其实开始了一个与深度学习相关的项目。

I actually start a little bit of deep learning related project.

那是我在滑铁卢大学攻读博士期间。

When I was doing my PhD study at the University of Waterloo.

好的。

Okay.

当时我正在研究我的论文，主题是计算机视觉，即用计算机检测物体，以便我们的机械臂能够抓取它。

At the time at that time I was study my thesis is computer vision, to use computer to detect object so that our robotic arm can pick it up.

所以不仅要识别物体，还要定位物体。

So not only recognize a project, the object, but also to locate the project.

当时，我们基本上必须手动详细设计如何提取物体的边缘和特征，然后使用某种场景描述语言。

And at that time, basically, we have to manually craft in detail how to extract object, corner, feature, and then we use some type of more scene description language.

我使用了一种称为属性图的方法来描述场景，以便完成计算机视觉任务。

I use the so called attribute graph to describe a scene so that we can do the computer vision task.

那时我觉得，天啊，这实在太难了。

And at that time I feel it's man, this is just so difficult.

是的。

Yeah.

那时我也尝试过神经网络。

And I tried the neural network at that time.

我当时尝试了两层的神经网络，想用它来做一些基础的特征检测，在大型机上训练了几个月，但根本没成功。

And I tried using about two layer neural network, try to just do some basic feature detection, trained a few months on mainframe computer, and this never really worked.

嗯哼。

Mhmm.

再快进到现在，我想大概是二十年后了。

And fast forward now, about I think about twenty years.

我惊讶地发现，我们现在训练的神经网络通常有几百万个参数。

And I was just amazed the our training, we train our neural network usually have a few million parameter.

通常至少有几千层，最多能达到一百层。

And usually like a at least a few thousand layer, up to a 100 layer.

是的。

Yeah.

我们现在基本上用几块NVIDIA显卡就能完成训练，最多一个月就能进行非常精细的训练。

And we basically using a couple of NVIDIA card to do the training, do a really refined training in in a month at most.

所以这不仅仅是速度的提升，更是根本性的变革。

So it's completely not only about the speed, it is the fundamental changes.

是的

Yeah.

我们正在与阿尔伯特·杨对话。

We're speaking with Albert Yang.

杨博士是Topaz Labs的创始人兼总裁，该公司以其广为人知且非常受欢迎的插件套件而闻名，这些插件用于编辑、增强和锐化照片，现在也支持视频素材。

Doctor Yang is founder and president of Topaz Labs, is a well known, really popular suite of plugins for editing, enhancing, sharpening, photographs, and also now, video footage as well.

你提到过，阿尔伯特，你在滑铁卢大学的博士研究工作。

You mentioned, Albert, your work, your PhD work at University of Waterloo.

如果你愿意的话，让我们稍微回溯一下，谈谈你从滑铁卢大学毕业后做了什么，是什么促使你创立了Topaz公司，以及你是如何对摄影和数字图像产生兴趣的。

Let's back up a little bit if you would and talk a little bit more about coming out of Waterloo, what you were doing, and then what led you to start Topaz, and also just how you got interested in photography and digital imagery.

我的研究兴趣始终集中在计算机视觉、图像和音频信号增强上。

My research interest is always in computer vision, image and audio signal enhancement.

在我的滑铁卢大学博士学习期间，我所在的实验室叫做PAMI，即专利分析与机器智能实验室。

And I basically during my PhD study at the University of Waterloo, the library I was in called the PAMI, Patent Analysis and Machine Intelligence.

因此，我当时在那里从事机器视觉方面的研究。

So I was there doing machine vision.

毕业后，我加入了几家小公司。

Afterward, after my study, I joined a couple of small company.

它们大多是国防承包商，从事图像、雷达和声呐信号处理。

Most of them are defense contractor doing image, radar, and the sonar signal processing.

然后我加入了另一家公司，从事语音、回声消除和降噪等信号增强工作。

And then I I joined the auto doing speech, echo cancellation, noise suppression, those type of also signal enhancement.

好的。

Okay.

正如你所知，硅谷很有吸引力，于是我去了圣何塞，加入了Techwell的创始团队，这家公司专注于数字视频增强半导体和集成电路。

And then as you know, the Silicon Valley appeal and I go to San Jose and join the founding team of Techwell, which is doing digital video enhancement semiconductor, IC.

好的。

Okay.

之后，我创办了另一家名为Forti Media的小公司，专注于微型麦克风阵列集成电路。

And after that, I founded another small company called Forti Media, which is doing small microphone array IC.

我们使用多个麦克风来拾取清晰的信号。

Basically, we use multiple microphone to pick up a clear signal.

现在，这些东西几乎无处不在。

Right now, those things are almost everywhere.

对。

Right.

你看，这个回声。

You see, the echo.

我们公司实际上是最早涉足这个方向的少数公司之一。

We we our company actually probably one of the very few to start this direction.

在一家公司上市后，我因为家庭原因回到了达拉斯。

And after one of the company go IPO, I come back to to Dallas due to family reason.

对。

Right.

我投资了Topaz Lab，因为我是个摄影爱好者。

And basically, funded Topaz Lab because I am a hobby photographer.

我用的一些工具并不太令人满意。

And some of the tool I use, they just are not very satisfactory.

所以我开始自己开发插件。

So I start doing plug in myself.

后来我把它们发布到网上，逐渐获得了一些关注。

And later on, just put to the Internet and they start getting certain traction.

是的。

Yeah.

这就是Topaz Lab的由来。

That's how Topaz Lab got wounded.

哦，太酷了。

Oh, very cool.

当你试图解决自己遇到的问题，而这个问题最终也帮到了其他人时，这总是很棒的。

It's always great when it's you're trying to solve a problem for yourself, and then it turns into something that other people can use as well.

我听你说话时在想，以你计算机视觉方面的背景，同时又从事音频信号处理、音频和语音降噪等工作，

I was wondering as you were speaking, you know, given your background in, you know, studying computer vision, obviously, but working with audio signal processing and and noise reduction in in audio and speech and everything.

那么，处理图像和处理声音之间有多相似或不同呢？

How similar or different is it working with images versus sound?

它们各自都有特定的特性。

They all have certain their own characteristics.

然而，总体而言，它们都可以使用类似的方法。

However, overall, they all can use similar methodology.

实际上，可以把它们看作非常相似。

Consider them very similar, actually.

深度学习在图像和音频信号处理领域如此成功的原因之一，是我们拥有海量的可用数据集。

The one of the thing deep learning is the reason deep learning is so successful in image and audio signal processing area is we have a vast amount of available dataset.

基本上，如果我去YouTube下载视频，我能获得无限多的数据。

Basically, if I go on YouTube and then start downloading, I have unlimited amount.

对吧？

Right?

但这一点在许多其他应用中并不成立。

And this is not true for many other applications.

例如，在金融领域或其他一些领域。

For example, financial area or some other area.

因此，深度学习在图像和音频领域的成果非常令人印象深刻。

That's why the result in image or audio area in term of deep learning are really impressive.

那么，Topaz 现在的团队规模有多大？

And so how big is the team at Topaz right now?

目前我们有大约20人。

Right now we have about 20 people.

好的。

Okay.

大部分工作是集中在研究与产品化相结合，还是

And is most of that focused on is it kind of a mix of research and then productization or is

主要是以产品为导向？

it mostly product focused?

我们公司整体上非常注重产品。

Our company is very product focused as a group.

大多数人是软件开发人员和机器学习研究人员。

Majority of people are software developer and machine learning researchers.

我甚至不敢称自己是个摄影爱好者，就凭我拍的照片质量来说。

I wouldn't even call myself a hobbyist photographer, just judging by the quality of my photos.

对我来说，这真的取决于相机本身的技术进步，而如今对像我这样的人来说，相机往往就是我的手机。

It really depends for me on, you know, the the advances that are happening in the cameras themselves, which these days for many people like me are are my phone.

我知道，当拍照时，设备上的软件会进行大量的处理。

And I do know that there's a lot of processing that happens, know, when the picture is being taken and in the software on the device itself.

所以我想知道，基于你的工作背景以及Topaz目前所做的工作，你认为这个行业将走向何方？

And so I'm wondering given, you know, your history of work and and the work that Topaz is doing right now, where do you see the the industry headed?

你可以具体谈谈Topaz，也可以更广泛地谈谈设备上发生的变化，比如多镜头和传感器如何协同生成一张图像，以及软件中发生的种种操作。

And feel free to answer specific to Topaz or kinda more broadly when it comes to kind of, know, what's happening on the devices and and multiple lenses and sensors, you know, combining to make a single image and so much happens in the software.

但你所处的这个领域——我不知道‘后期处理’这个说法是否准确——但你所做的一切，都是在拍摄完成后对图像进行编辑。

But then this whole world that you're in with, I don't know if post processing is the right term but everything that you're doing you know after the fact to edit images.

总的来说，深度学习和人工智能正在如何推动我们所熟知的摄影流程？

Where in general is deep learning and AI kind of pushing the whole process of photography as we know it?

据我所见，深度学习正在渗透到每一个领域。

From what I can see, deep learning is penetrating every area.

在设备端，你已经可以看到很多进展，比如谷歌推出的夜景模式，它利用深度学习来实现正确的效果。

On the device side, you can already see a lot of progress as Google put in the poultry mode, which is using deep learning to do Right.

是的。

Right.

来模糊背景。

To do blur the background.

对。

Yep.

我认为这些变化还会继续发生。

And those things I see are going to continue to happen.

特别是出现了一个新领域，叫做计算摄影。

And especially there's a new field called computational photography.

他们甚至可以利用光场来合成成像过程，从而取得非常惊人的效果。

They basically, they can even use the light field to synthesize the imaging process to achieve a pretty amazing result.

所以这些变化还会持续发生。

So those things continue to happen.

然而，与后期处理环境（如PC、Mac甚至服务器端）相比，设备的计算能力始终有限。

However, the device has always much limited computation power compared with the post processing environment, PC or Mac or even on the server side.

因此，我也看到在后期处理方面，尤其是在桌面端，还有巨大的发展空间。

So I can also see there's a huge room on the post processing, especially on the desktop side.

另一个有趣的方面是，我发现直到最近，基于桌面的深度学习应用仍然非常罕见，因为实际上开发起来非常困难。

This is another interesting aspect is I find until recently desktop deep learning based app is very rare because it is actually very hard to develop.

在PC或Mac上开发缺乏良好的基础设施。

There's not good infrastructure to develop on PC or Mac.

它们有一些库，但没有一个是非常适合生产环境的。

They have a have a few library, not none of them are very production ready.

而且存在很多兼容性问题。

And there are a lot of compatibility issue.

因此，人们通常首先使用云解决方案来开发他们的深度学习方案。

So people usually first develop their deep learning solution using cloud solution.

是的。

Right.

但我们发现大多数客户都非常讨厌它。

But we find the most customers really hate it.

我们最初把视频增强解决方案放在云端，但根本没人用。

We actually have our video enhancement solution first on the cloud and nobody use it.

几乎没人用。

Very few people use it.

所以我看到在PC和微型设备的后期处理方面，有很多机会。

So I can see in the post processing side on the PC and the mic device, I see a lot of opportunity.

当然，在云端的这些领域也有巨大的机会。

Of course, on the cloud side, is also tremendous opportunity in those area.

你

Do you

你平时自己拍照时还会用老式的胶片相机吗？

ever shoot on old fashioned film when you're doing your own photography?

不再用了。

Not anymore.

不再了。

Not anymore.

我其实猜到你会这么回答，但我也不确定。

I I kind of figured that was gonna be the answer, but but I I don't know.

我自己对老科技还挺怀旧的，所以才问一下。

Get nostalgic for old tech myself, so I I had to ask.

是的。

Yeah.

我认为，机会非常有趣，尤其是当你把现在的手机摄像头和五年前的大型昂贵相机相比，特别是在低光环境下的表现时。

I think the again, the opportunity is is very interesting to see if you compare with the current phone camera as a let's say low lighting performance compare with even just five years ago, a bigger camera, a very expensive camera.

手机摄像头在低光环境下的表现已经和过去的大型相机相当了。

The phone camera's low lighting performance is very comparable with the old big camera.

对。

Right.

而这主要是因为边缘端的集成。

And this is actually mainly because of the edge level integration.

也许不是深度学习，但一定的计算能力可以完成大量处理。

Maybe not deep learning, but a certain computation power, they can do a lot of processing.

这些进步仍在持续发生。

Those are continuing to happen.

因此，设备端变得越来越好。

So the device side is getting better and better.

然而，在后处理方面，计算能力始终会快上几个数量级。

However, on the post processing side, the computing power, it will always be a few factor faster.

对。

Right.

有了更强的算力，我们可以做得好得多。

And with that more power, we can do much much better job.

这就是为什么我们认为后处理领域还有很长的路要走。

That's why we feel the post processing area still have a long time to go.

是的。

Yeah.

你一直在计算机视觉和信号处理领域工作，已经有一段时间了。

You've you've been working in in computer vision, you know, for for a while now in signal processing.

如果让我请你回顾一下，有没有哪些时刻，或者特别的挑战或突破，让你感到特别惊讶？

You know, if I could ask you to kinda look back, are there any moments or any particular, you know, challenges or breakthroughs that that stand out to you as particularly surprising?

不管是不是直接与图像处理相关，但只要是与计算机视觉、深度学习这些领域有关的，任何让你没想到却成为重大转折点的事情，都可以说说。

And whether they have to do specifically with image processing or not, But related to, you know, computer vision, deep learning, that whole thing, anything that stands out at you is something that that really you weren't expecting but turned out to be a big moment.

我想再次回到计算能力的提升，这主要归功于GPU。

I think one, again, I go back to the computing power improvement, largely due to GPU.

这基本上是一个颠覆性的改变。

This is basically a co game changer.

如果没有它，几乎是不可能实现的。

Without it, it is pretty much impossible.

这对于一家小公司来说。

It's basically for a small company.

不像大公司那样，可以拥有成千上万台服务器连接在一起，开展像样的深度学习项目。

Unlike a big company, they can have a server farm with thousands of server connected and do some decent deep learning project.

像我们这样的小公司，根本不可能训练出一个像样的图像增强神经网络。

A small company like us, there's no way we can train a reasonable image enhancement neural network.

例如，如果我现在只用一台普通的电脑而不使用GPU，训练一个网络可能需要好几年。

For example, if I now I just use a good computer without using GPU, training one network will probably take a couple of years Mhmm.

所以这确实是一个颠覆性的改变。

If you're secure at So a that is actually a game changer.

是的。

Yeah.

当然。

Absolutely.

而且，我们的大多数用户，他们的台式机都配备了一定类型的GPU。

And also, majority of the our our user, the desktop have some type of GPU.

对。

Right.

稍微高端一点的。

A little higher end.

这也是一个重要的推动因素。

That is also a big enabler.

所以现在，我不再让你回顾过去，而是请你展望一下未来，因为这是我们节目对嘉宾的常规做法。

And so now instead of asking you to look back, I'll ask you to look ahead a little bit because this is what we do to our guests on the show.

你有没有预见到什么？也许是正在研究的东西，或者未来可能实现的突破，你认为这些会如何塑造未来两到三年，甚至五年的数字摄影和视频领域？

Is there anything that that you're kind of foreseeing, maybe something you're working on or maybe something you're anticipating down the line that you think is gonna shape, let's say, next two, three, even five years of of digital photography and video?

从近期来看，我看到研究领域中许多非常有趣的进展将持续被转化为产品。

In the near term, from what I can see is basically many of the really interesting development in research area will be continued, will be productized continually.

我们的Topaz实验室将是专注于将新成果推向客户的企业之一。

Our Topaz Lab will be one of the company really focusing on introducing some new development to our customer.

而且，即使是现有的产品类别，比如图像放大、降噪和去锯齿处理等，也还有大量潜力。

And also, even for existing product category like image enlargement, noise reduction and the deburrer masking, those type of deal.

我们希望将许多新的技术进展融入产品中，让它们变得更好。

There are so many potential and the new development that we would like to incorporate into our product to make it even better.

例如，目前的智能水平还不够聪明。

For example, right now the intelligence is still not very smart.

人们需要为不同类型的图像问题选择不同的模型。

People have to select a different model for different type of image problem and so on.

所以我们正在努力让它变得更智能。

So we are trying to make it smarter.

例如，一个更大的神经网络，能够适应并识别你的图像存在什么问题，并希望自动将其修正到最佳状态。

For example, one bigger neural network that can adapt it, discover what type of problem your image have and hopefully automatically correct it to the best.

这正是我需要的。

That's what I need.

我只要点一下那个魔法棒按钮，让它自动处理，让我的孩子在照片里看起来更好。

I I hit that magic wand button and just let it let it do its thing to make my kids look better in the pictures.

是的。

Yeah.

它即将到来。

It is coming.

这项技术正朝着这个方向发展，并且已有新的方法在这一领域取得进展。

The technology is developing in this direction and there are new method working in this area.

嗯，我个人非常期待，迫不及待想看看接下来会有什么新进展。

Well, I for one look forward to it and can't wait to see what's coming next.

我知道你们的网站 topazlabs.com 提供了大量产品的免费试用、演示等内容，但我还听说你们有一个相当不错的博客。

I know your website topazlabs.com has places there free trials of a lot of the products and demonstrations and such, but I also hear you've got a pretty good blog.

我个人确实如此。

I personally, yeah.

我偶尔会在我们的博客上写几篇文章，介绍我们开发背后的背景信息，试图揭开人工智能技术的神秘面纱，并将其放在一定的历史背景下加以说明。

I I did occasionally write a couple of article in our blog site to describe the background information about our development and and just try to demystify the AI technology and put it into a little bit historical perspective.

我真的很享受撰写这些文章的过程。

I really enjoyed writing those articles.

太好了。

Excellent.

那就是 topazlabs.com/blog。

Well that is at topazlabs.com/blog.

在网站上很容易就能找到，对吧。

Easy enough to find right there on the site.

除了Topaz Labs网站之外，还有其他地方可以推荐人们去了解更多信息吗？

Anywhere else you would point people to go if they want to learn more or is it all on the Topaz Labs site?

是的。

Yeah.

大部分信息都在Topaz网站上。

Majority of the information is on Topaz website.

如果你在Google上搜索Topaz AI，也会找到很多信息。

And also if you Google Topaz AI, you'll find a lot of information.

杨博士，感谢您抽出时间，网站上有很多内容值得探索。

Well, Doctor Yang, we appreciate the time and there's a lot of stuff to play with on the website.

正如我所说，我确实能用得上这些工具。

I certainly, as I've said, I can use it.

显然，还有许多其他爱好者和专业人士也在使用。

And clearly, there are lots of others out there, both hobbyists and professionals who do as well.

感谢您为推动这一领域发展所做的一切。

So thanks for all you're doing to advance the field.

非常感谢。

Thank you very much.

真的很感激。

Really appreciate.

非常享受这次访谈。

Really enjoy your interview.

这是我的荣幸。

It's my pleasure.