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

用神经网络重新聚焦视频 – 译学馆
未登录,请登录后再发表信息
最新评论 (0)
播放视频

用神经网络重新聚焦视频

Refocusing Videos With Neural Networks | Two Minute Papers #173

亲爱的学霸们 这是Károly Zsolnai-Fehér带来的两分钟论文
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
每当我们用相机拍完照 事后查看时 我们常常觉得
Whenever we take an image with our camera, and look at it after an event, we often feel
很多照片都离完美差那么点 要是照片
that many of them are close to perfect, if only
不那么模糊 或者焦距能再远一点就好了
it was less blurry, or the focus distance was a bit further away.
但那个神奇的时刻现已过去 向朋友们展示照片时
But the magic moment is now gone, and there is nothing to do other than cursing at the
除了诅咒那些留下的模糊画面 我们别无办法
blurry footage that we’re left with when showing it to our friends.
然而 如果能利用光场 我们就能在
However, if we have access to light fields, we can
拍完照后对一些相机参数作出改变
change some camera parameters after the photo was taken.
这包括改变焦距
This includes changing the focal distance,
或者甚至微调相机的视角
or even slightly adjusting the viewpoint of the camera.
那多酷啊!
How cool is that!
这可以通过光场相机来完成
This can be accomplished by a light field camera
它也被称为全光相机
which is also referred to as a plenoptic camera.
它不仅记录光强度 而且也记录入射光的方向
This tries to record not only light intensities, but the direction of incoming light as well.
早期这通常是用一系列摄像机来实现的
Earlier, this was typically achieved by using an array of cameras.
它们既昂贵又笨重
That’s both expensive and cumbersome.
当只用一个光场相机时会出现一个问题:
And here comes the problem with using only one light field camera: because of the increased
由于需要记录的数据量增加
amount of data that they have to record,
而现在的光场相机每秒只能拍摄3帧
current light field cameras are only able to take 3 frames per second.
如果我们想对视频进行这样的后期编辑 那它就很难令人满意
That’s hardly satisfying if we wish to do this sort of post-editing for videos.
通过为光场相机附加一个标准相机 这一工作
This work offers a novel technique to remedy this situation by attaching
为修正这一情况提供了一项新技术
a standard camera to this light field camera.
目标是 标准相机每秒可以到30帧
The goal is that the standard camera has 30, so tons of frames per second, but with little
但附加信息很少 而光场相机每秒只有几帧
additional information, and the light field camera, which has only a few frames per second,
但有大量的附加信息
but, with a ton of additional information.
如果我们以一种聪明的方式将所有信息结合起来
If we stitch all this information together in a smart way,
也许就是对视频进行充分的光场编辑的一种可能方法
maybe it is a possibility to get full light field editing for videos.
早些时候 我们谈到了可以填补
Earlier, we have talked about interpolation techniques that
视频中缺失帧的插值技术
can fill some of the missing frames in videos.
这样 一个镜头里我们也许能每隔一帧填充一次
This way, we can fill in maybe every other frame in a footage,
或许我们能比那样更大方一点
or we can be a bit more generous than that.
但是 如果一秒显示3帧
However, if we’re shown 3 frames a second,
为了创建一个平滑的视频
and we have to create a smooth video
我们就必须让算法像是
by filling the blanks would almost be like
从漫画书创建电影那样填充空白
asking an algorithm to create a movie from a comic book.
那会很棒 但我们还不到那一步
This would be awesome, but we’re not there yet.
缺失信息太多了
Too much information is missing.
拼接过程需要比这多一点的信息 关键想法是
This stitching process works with a bit more information than this, and the key idea is
使用两个卷积神经网络来填充空白:一个用来
to use two convolutional neural networks to fill in the blanks: one is used to predict
预测流动 它描述场景中物体的动作和旋转
flows, which describe the movements and rotations of the objects in the scene,
另一个预测物体的最终外观
and one to predict the final appearance of the objects.
基本上 一个用于物体的移动 一个用于物体的形状
Basically, one for how they move, and one for how they look.
结果绝对令人难以置信
And the results are just absolutely incredible.
它同样非常迅速 用不到十分之一秒
It is also blazing fast and takes less than
的时间就能创建这些新视图中的一个
a tenth of a second to create one of these new views.
这里 你能看到最终的程序是如何改变视频中
Here, you can see how the final program is able to change the focal distance
任意帧的焦距的 或者我们甚至可以
of any of the frames in our video, or we can even
点击图像中的某一物体来对焦
click on something in the image to get it in focus.
所有这一切都是在视频拍摄后完成的
And all this is done after the video has been taken.
该项目的源代码也是可用的
The source code of this project is also available.
随着更多的改进 这将在电影业中非常有用
With some more improvements, this could be tremendously useful in the film industry,
因为导演可以在拍摄后调整自己的场景 而不仅仅是叹息
because the directors could adjust their scenes after the shooting, and not just sigh over
拍摄的不准确和错失良机
the inaccuracies and missed opportunities.
这只是其他许多可能的应用中的一个
And this is just one of many possible other applications.
绝对很棒
Absolutely amazing.
如果你喜欢这一期 别忘了订阅两分钟论文
If you enjoyed this episode, don’t forget to subscribe to Two Minute Papers
同时确保点击铃形图标 从而不错过一期节目
and also make sure to click the bell icon to never miss an episode.
谢谢您的收看和慷慨支持
Thanks for watching and for your generous support,
我们下次再见!
and I’ll see you next time!

发表评论

译制信息
视频概述

人们常常为拍摄后的对焦不准确而感到可惜,本期两分钟论文就为您介绍一个可以用来重新聚焦视频的神经网络。

听录译者

收集自网络

翻译译者

B11101001

审核员

审核团1024

视频来源

https://www.youtube.com/watch?v=EGnbAgbRIh4

相关推荐