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

Shape2vec:使用人工智能来理解3D形状 – 译学馆
未登陆,请登陆后再发表信息
最新评论 (0)
播放视频

Shape2vec:使用人工智能来理解3D形状

Shape2vec: Understanding 3D Shapes With AI | Two Minute Papers

学霸们大家好 这里是由Károly Zsolnai-Fehér带来的两分钟papers
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
这回我们讲的是一篇碉堡了的文章
This one is going to be absolutely amazing.
这项工作的目标在于帮助机器更好地理解
This piece of work is aimed to help a machine build a better understanding of images and
图像和3D几何形状
3D geometry.
试想我们有一个很大的图像和几何形状数据库 然后我们可以搜索
Imagine that we have a large database with these geometries and images, and we can search
数据库中的模型并和任意的输入输出进行比较
and compare them with arbitrary inputs and outputs.
这是啥意思呢?
What does this mean exactly?
举个栗子 它能以一段文字作为输入 像是“学校大巴” 然后自动检索
For instance, it can handle a text input, such as school bus and automatically retrieve
对应于这种文字描述的3D模型 草图和图像
3D models, sketches and images that depict these kinds of objects.
这听起来很棒 不过我们说过这个系统支持任意的输入和输出 意思是
This is great, but we said that it supports arbitrary inputs and outputs, which means
我们可以以一个椅子的3D几何形状作为输入 得到另一些有相似外观的
that we can use the 3D geometry of a chair as an input, and obtain other, similar looking
椅子(来自数据库中的)
chairs from the database.
这种技术真是了不起 它甚至支持以草图做输入 然后输出非常
This technique is so crazy, it can even take a sketch as an input and provide excellent
高质量的结果
quality outputs.
我们甚至可以给它一幅深度图作为输入 并且得到相当可靠的结果
We can even give it a heatmap of the input and expect quite reasonable results.
通常 这些图像和3D几何形状包含很多信息 如果想要
Typically, these images and 3D geometries contain a lot of information, and to be able
通过这些信息比较哪个和哪个数据更相似 我们就需要把这些信息压缩成更加简洁的
to compare which is similar to which, we have to compress this information into a more concise
描述形式
description.
这些描述信息提供了一个一致的比较标准
This description offers a common ground for comparisons.
我们把这一套方法叫做嵌入技术
We like to call these embedding techniques.
这里 大家可以看到一个文字类嵌入数据的二维可视化
Here, you can see an example of a 2D visualization of such an embedding of word classes.
从数据库中检索时 系统会压缩用户的输入数据并映射到
The retrieval from the database happens by compressing the user-provided input and putting
这个嵌入空间内 然后提取嵌入空间中距离输入数据最近的结果
it into this space, and fetching the results that are the closest to it in this embedding.
在强大的机器学习算法兴起之前 这种嵌入数据通常是
Before the emergence of powerful learning algorithms, these embeddings were typically
手动标注的
done by hand.
但是现在 我们有了深度神经网络 它可以自动创建我们需要的结果
But now, we have these deep neural networks that are able to automatically create solutions
这些结果在某种程度上是最优的 意味在一定的规则下 它们
for us, that are in some sense, optimal, meaning that according to a set of rules, it will
总是好于手工标注
always do better than we would by hand.
只需要开着电脑让它跑一晚上 然后我们去睡觉 就能得到更好的结果
We get better results by going to sleep and leaving the computer on overnight than we
如果交给人日夜不停地去标注 即使用最先进的算法 也得10年
would have working all night using the finest algorithms from ten years ago.
是不是听起来很不可思议?
Isn’t this incredible?
有趣的是 我们可以用这种方法来处理不同的数据类型:
The interesting thing is that here, we are able to do this for several different representations:
比如 一个3D几何形状 或者一幅2D彩色图片 又或者一个简单的词 都可以被嵌入
for instance, a piece of 3D geometry, or 2D color image, or a simple word, is being embedded
到同一个向量空间中 而这也使得这种令人激动的
into the very same vector space, opening up the possibility of doing these amazing comparisons
跨类别比较成为了可能
between completely different representations.
结果本身就证明了这项研究的价值
The results speak for themselves.
这又是一个卷积神经网络的坚实佐证 你可以看到
This is another great testament to the power of convolutional neural networks and as you
人工智能和机器学习的发展速度着实令人惊艳
can see, the rate of progress in AI and machine learning research is absolutely stunning.
这里 我要给那些发现了我们的新结尾曲的学霸们竖个大拇指
Also big thumbs up for the observant Fellow Scholars out there who noticed the new outro
当然 还有一些别的小改动
music, and, some other minor changes in the series.
如果你也看出来了 那你真的是我们两分钟papers的铁杆粉了!
If you are among those people, you can consider yourself a hardcore Two Minute Papers Scholar!
来击个掌吧伙计
High five!
感谢收看和捧场 下次见咯
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

这次up主介绍了一篇来自SIGGRAPH Asia 2016的文章,这篇文章通过深度学习的方法实现了对任意输入的3D形状检索,十分令人惊艳

听录译者

收集自网络

翻译译者

GraphiCon-origamidance

审核员

知易行难

视频来源

https://www.youtube.com/watch?v=bB54Wz4kq0E

相关推荐