Blaise Aguera y Arcas: Jaw-dropping Photosynth demo

46,128 views ・ 2007-06-26

TED


請雙擊下方英文字幕播放視頻。

譯者: Wang Qian 審譯者: Bill Hsiung
00:25
What I'm going to show you first, as quickly as I can,
0
25000
2548
首先,我要用最快的速度為大家示範
00:27
is some foundational work, some new technology
1
27572
3769
一些新技術的基礎研究成果。
00:31
that we brought to Microsoft as part of an acquisition
2
31365
2611
正好是一年前, 微軟收購了我們公司,
00:34
almost exactly a year ago.
3
34000
1821
而我們為微軟帶來了這項技術,它就是「海龍」(Seadragon)。
00:35
This is Seadragon, and it's an environment
4
35845
2368
「海龍」是一個軟體環境,你可以通過它以近景或遠景的方式
00:38
in which you can either locally or remotely interact
5
38237
2476
00:40
with vast amounts of visual data.
6
40737
2119
流覽浩瀚的視覺化資料。
00:43
We're looking at many, many gigabytes of digital photos here
7
43165
3404
我們在這裏看到的是,非常多千兆位元組的數位像片,
00:46
and kind of seamlessly and continuously zooming in,
8
46593
2915
我們可以對它們可以進行持續並且平滑的放大,
00:49
panning through it, rearranging it in any way we want.
9
49532
2545
可以通過全景的方式流覽它們,還可以對它們進行重新排列。
00:52
And it doesn't matter how much information we're looking at,
10
52389
3587
不管所見到的資料有多少、
00:56
how big these collections are or how big the images are.
11
56000
2976
圖像集有多大或是圖像本身有多大。
00:59
Most of them are ordinary digital camera photos,
12
59000
2286
以上展示的圖片,大部分是來自一般數位相機的照片,
01:01
but this one, for example, is a scan from the Library of Congress,
13
61310
3144
但這個例子不同,它是一張來自國會圖書館的掃描圖片,
01:04
and it's in the 300 megapixel range.
14
64478
2818
擁有3億個像素。
01:07
It doesn't make any difference
15
67320
1656
然而,這沒有造成任何不同,
01:09
because the only thing that ought to limit the performance of a system like this one
16
69000
4144
因為能夠限制像這樣的系統效能的唯一因素,
是你所使用的螢幕所正在顯示的像素數量。
01:13
is the number of pixels on your screen at any given moment.
17
73168
2777
01:15
It's also very flexible architecture.
18
75969
1970
「海龍」同時也是一個非常靈活的架構。
01:17
This is an entire book, so this is an example of non-image data.
19
77963
3727
這是一本完整的書,是非圖形式資料的一個例子。
01:21
This is "Bleak House" by Dickens.
20
81714
2787
這是狄更斯所著的《荒涼山莊》,每一欄就是一個章節。
01:24
Every column is a chapter.
21
84525
2784
01:27
To prove to you that it's really text, and not an image,
22
87333
3643
為了向你們證明這真的是文字而非圖片,
01:31
we can do something like so, to really show
23
91000
2048
我們可以這樣操作,
01:33
that this is a real representation of the text; it's not a picture.
24
93072
3192
大家可以看得出來這真的是文字,而不是一張圖片。
01:36
Maybe this is an artificial way to read an e-book.
25
96288
2664
也許這會是一種閱讀電子書的方式,
01:38
I wouldn't recommend it.
26
98976
1200
但是我個人不會推薦這麼做。
01:40
This is a more realistic case, an issue of The Guardian.
27
100200
2848
接下來是一個更加實際的例子,這是一期《衛報》。
01:43
Every large image is the beginning of a section.
28
103072
2286
你所看到的每一張大圖片,就是各版頭條,
01:45
And this really gives you the joy and the good experience
29
105382
2904
而報紙或雜誌的紙本,本身就包含了各種比例的媒材,
01:48
of reading the real paper version of a magazine or a newspaper,
30
108310
5183
因此這樣閱讀的時候,讀者會得到更好的閱讀體驗,
01:53
which is an inherently multi-scale kind of medium.
31
113517
2435
從而享受閱讀的樂趣。
01:55
We've done something
32
115976
1000
我們在這裏做了點小小的更動,
01:57
with the corner of this particular issue of The Guardian.
33
117000
2976
在這一期《衛報》的一角。
02:00
We've made up a fake ad that's very high resolution --
34
120000
2976
我們刊登了一個非常高解析度的虛構廣告 —
02:03
much higher than in an ordinary ad --
35
123000
2198
比你平常看到的普通廣告的解析度要高很多,
02:05
and we've embedded extra content.
36
125222
1754
我們並在圖片中嵌入了額外的內容。
02:07
If you want to see the features of this car, you can see it here.
37
127000
3048
如果你希望看到這輛車的特性,你可以看這裏,
02:10
Or other models, or even technical specifications.
38
130072
4180
你還能看到其他的型號,甚至技術規格。
02:14
And this really gets at some of these ideas
39
134276
3315
這種方式在一定程度上,
02:17
about really doing away with those limits on screen real estate.
40
137615
4661
避開了螢幕面積的限制。
02:22
We hope that this means no more pop-ups
41
142300
2111
我們希望這個技術能夠減少不必要的彈出視窗,
02:24
and other rubbish like that -- shouldn't be necessary.
42
144435
2541
及其他類似的,不必要的垃圾。
02:27
Of course, mapping is one of those obvious applications
43
147000
2658
當然,對於像這樣的技術,
02:29
for a technology like this.
44
149682
1294
數位地圖也是顯而易見的應用之一。
02:31
And this one I really won't spend any time on,
45
151000
2191
對此,我真的不想花費太多的時間進行介紹,
02:33
except to say that we have things to contribute to this field as well.
46
153215
3334
我只想告訴大家,我們對這個領域也貢獻了一己之力。
02:37
But those are all the roads in the U.S.
47
157213
1858
這些是將美國的所有道路,
02:39
superimposed on top of a NASA geospatial image.
48
159095
4565
疊加在太空總署的地理空間影像上。
02:44
So let's pull up, now, something else.
49
164000
1976
現在,我們先放下這些,看看其他的。
02:46
This is actually live on the Web now; you can go check it out.
50
166000
2976
實際上,這項技術已經放到網路上了,大家可以自己去體驗一下。
02:49
This is a project called Photosynth, which marries two different technologies.
51
169000
3704
這個計畫名叫「相片合成」 (Photosynth),
它實際上融合了兩個不同的技術:
02:52
One of them is Seadragon
52
172728
1248
一個是「海龍」,
02:54
and the other is some very beautiful computer-vision research
53
174000
2906
而另一個則是源自華盛頓大學的研究生 Noah Snavely,
02:56
done by Noah Snavely, a graduate student at the University of Washington,
54
176930
3462
所進行的電腦視覺化研究的美麗成果。
03:00
co-advised by Steve Seitz at U.W.
55
180416
1829
這項研究還得到了華盛頓大學 Steve Seitz
03:02
and Rick Szeliski at Microsoft Research.
56
182269
1978
和微軟研究中心 Rick Szeliski 的共同指導。這是一個非常漂亮的合作成果。
03:04
A very nice collaboration.
57
184271
1733
03:06
And so this is live on the Web. It's powered by Seadragon.
58
186412
3108
現在各位看到的是我們連上網路的即時示範,它是根基於「海龍」技術。
03:09
You can see that when we do these sorts of views,
59
189544
2504
你可以看到,我們輕鬆地對圖片進行多種方式的查看,
03:12
where we can dive through images
60
192072
1723
就好像潛入這些影像一般,
03:13
and have this kind of multi-resolution experience.
61
193819
2334
擁有了這種多解析度的瀏覽體驗。
03:16
But the spatial arrangement of the images here is actually meaningful.
62
196177
3799
不過,在這邊,這些圖片空間上的關係事實上是有意義的。
03:20
The computer vision algorithms have registered these images together
63
200000
3191
電腦視覺演算法將這些圖片聯繫到一起,
03:23
so that they correspond to the real space in which these shots --
64
203215
3761
那麼這些圖片就能將真實空間給呈現出來了,
03:27
all taken near Grassi Lakes in the Canadian Rockies --
65
207000
3300
而我們正是在這個地方拍了上述的照片 — 這些照片都是在
03:30
all these shots were taken.
66
210324
1663
加拿大洛磯山脈的格拉西湖附近拍下的 — 所有照片都是在這裏拍下的。
03:32
So you see elements here
67
212011
1467
03:33
of stabilized slide-show or panoramic imaging,
68
213502
6013
因此,這邊你可以看到穩定幻燈片播放的元素或者環景影像,
03:39
and these things have all been related spatially.
69
219539
2437
而這些內容在空間上都是互相關聯的。
03:42
I'm not sure if I have time to show you any other environments.
70
222000
3000
我不確定我是否有時間為你們示範其他環境的例子。
03:45
Some are much more spatial.
71
225024
1431
有些其他例子比這個的空間感還要強。
03:46
I would like to jump straight to one of Noah's original data-sets --
72
226479
3945
下面讓我們來看一下去年夏天,
03:50
this is from an early prototype that we first got working this summer --
73
230448
3552
Noah 早期所建立的資料集之一,
這是來自於「相片合成」技術早期的原型階段。
03:54
to show you what I think
74
234024
1894
我認為,
03:55
is really the punch line behind the Photosynth technology,
75
235942
3838
這是我們這項技術最搶眼之處。
03:59
It's not necessarily so apparent
76
239804
1561
「相片合成」技術不單單像我們剛剛在
04:01
from looking at the environments we've put up on the website.
77
241389
2895
網站上所示範的環境般,那麼的簡單明瞭。
04:04
We had to worry about the lawyers and so on.
78
244308
2177
主要因為我們製作網站時,要顧慮很多法律問題。
04:06
This is a reconstruction of Notre Dame Cathedral
79
246509
2301
這裡是利用 Flickr 網站上
04:08
that was done entirely computationally from images scraped from Flickr.
80
248834
3457
的照片,並完全以電腦重建的巴黎聖母院。
你所要做的只是在 Flickr 網站上輸入「巴黎聖母院」,
04:12
You just type Notre Dame into Flickr,
81
252315
2019
04:14
and you get some pictures of guys in T-shirts, and of the campus and so on.
82
254358
3854
然後便能看到很多照片,包括在那邊留影的遊客等等。
每一個橘色的錐形都代表了一張
04:18
And each of these orange cones represents an image
83
258236
3146
04:21
that was discovered to belong to this model.
84
261406
3234
用來建立模型的照片。
04:26
And so these are all Flickr images,
85
266000
1976
這些全部是來自 Flickr 的圖片,
04:28
and they've all been related spatially in this way.
86
268000
2976
被這樣在空間裡被串聯起來。
04:31
We can just navigate in this very simple way.
87
271000
2334
接著,我們便可如此自然的進行瀏覽。
04:35
(Applause)
88
275000
3920
(掌聲)
04:42
(Applause ends)
89
282557
1014
04:43
You know, I never thought that I'd end up working at Microsoft.
90
283595
2954
說實話,我從來沒想過我會為微軟工作,
04:46
It's very gratifying to have this kind of reception here.
91
286573
3000
這樣受到歡迎,真挺令人高興的。
04:49
(Laughter)
92
289597
3379
(笑聲)
04:53
I guess you can see this is lots of different types of cameras:
93
293000
5048
我想你們可以看出,
這些照片來自很多不同的相機:
04:58
it's everything from cell-phone cameras to professional SLRs,
94
298072
3161
從手機鏡頭到專業的單眼相機。
05:01
quite a large number of them, stitched together in this environment.
95
301257
3191
如此大量的不同品質的照片,全被在這個環境下
拼湊在一 起。
05:04
If I can find some of the sort of weird ones --
96
304472
2632
讓我來找些比較詭異的照片。
05:08
So many of them are occluded by faces, and so on.
97
308000
3322
看,不少照片包含了遊客的大頭照等等。
05:12
Somewhere in here there is actually a series of photographs -- here we go.
98
312595
4277
我記得這裡應該有
一系列的照片 — 啊,在這兒。
05:16
This is actually a poster of Notre Dame that registered correctly.
99
316896
3301
這實際上是一張有巴黎聖母院照片的海報,
05:20
We can dive in from the poster
100
320221
3216
我們可以鑽到海報裡,
05:23
to a physical view of this environment.
101
323461
3810
去看整個重建的環境。
05:31
What the point here really is
102
331421
1866
這裏的重點呢?便是我們可以
05:33
is that we can do things with the social environment.
103
333311
2591
有效地利用網路社群。我們可以從每個人那裡得到資料,
05:35
This is now taking data from everybody --
104
335926
3002
05:38
from the entire collective memory, visually, of what the Earth looks like --
105
338952
3871
將每個人對不同環境
的視覺記憶蒐集在一起,
05:42
and link all of that together.
106
342847
1749
並將它們連結起來。
05:44
Those photos become linked, and they make something emergent
107
344620
2839
當所有這些圖片交織在一起時,
所衍生出的東西,
05:47
that's greater than the sum of the parts.
108
347483
1953
要遠遠超過各部件的總和,
05:49
You have a model that emerges of the entire Earth.
109
349460
2356
這個模型所衍生出的,是整個地球。
05:51
Think of this as the long tail to Stephen Lawler's Virtual Earth work.
110
351840
4077
將之想像是 Stephen Lawler《虛擬地球》的長尾市場。(Stephen Lawler 是微軟「虛擬地球」專案主管)
05:55
And this is something that grows in complexity as people use it,
111
355941
3200
這類模型, 會隨著人們的
使用而不斷變得更複雜,
05:59
and whose benefits become greater to the users as they use it.
112
359165
3811
變得更加有價值。
06:03
Their own photos are getting tagged with meta-data that somebody else entered.
113
363000
3692
用戶的照片,會被其他人
輸入標注資料。
06:06
If somebody bothered to tag all of these saints
114
366716
3360
如果有人願意,為聖母院裡的所有聖賢輸入標注,
06:10
and say who they all are, then my photo of Notre Dame Cathedral
115
370100
2953
表明他們是誰,那我們聖母院的照片便會
06:13
suddenly gets enriched with all of that data,
116
373077
2098
一下子增加了許多資訊,
06:15
and I can use it as an entry point to dive into that space,
117
375199
2777
然後呢,我們便能以這張照片為起點,進入這個空間,
06:18
into that meta-verse, using everybody else's photos,
118
378000
2681
這個由很多人的照片所搭建的虛擬世界,
06:20
and do a kind of a cross-modal
119
380705
3301
從而得到一種跨越模型,
06:24
and cross-user social experience that way.
120
384030
3751
跨越用戶的社交體驗。
06:27
And of course, a by-product of all of that is immensely rich virtual models
121
387805
4171
當然了,這一切所帶來另外一個寶貴產物便是,
我們擁有地球上每一個有趣的地方,
06:32
of every interesting part of the Earth,
122
392000
1968
非常豐富的模型。
06:33
collected not just from overhead flights and from satellite images
123
393992
4487
這些模型的資料來源,不再僅限於空拍或衛星照片等等,
06:38
and so on, but from the collective memory.
124
398503
2052
而是來自全人類的集合記憶。
06:40
Thank you so much.
125
400579
1094
非常感謝!
06:41
(Applause)
126
401697
6863
(掌聲)
06:51
(Applause ends)
127
411967
1001
06:52
Chris Anderson: Do I understand this right?
128
412992
2326
Chris Anderson: 如果我理解正確的話,你們的這個軟體將能夠
06:55
What your software is going to allow,
129
415342
2497
06:57
is that at some point, really within the next few years,
130
417863
3476
在未來的幾年內,
07:01
all the pictures that are shared by anyone across the world
131
421363
4235
將來自全球網路使用者所共享的照片
07:05
are going to link together?
132
425622
1561
結合在一起?
07:07
BAA: Yes. What this is really doing is discovering,
133
427207
2387
BAA:是的。這個軟體的真正意義便是去探索,
07:09
creating hyperlinks, if you will, between images.
134
429618
2358
它在照片間建立超鏈結。
07:12
It's doing that based on the content inside the images.
135
432000
2584
這個結合的過程,
完全是基於照片的內容。
07:14
And that gets really exciting when you think about the richness
136
434608
3022
更令人興奮的
07:17
of the semantic information a lot of images have.
137
437654
2304
在於照片所包含的大量文字語義資訊。
07:19
Like when you do a web search for images,
138
439982
1960
譬如,你在網路上搜尋一張照片,
07:21
you type in phrases,
139
441966
1245
鍵入關鍵字後,網頁上的文字內容
07:23
and the text on the web page is carrying a lot of information
140
443235
2900
將包含大量與這個照片相關的資訊。
07:26
about what that picture is of.
141
446159
1502
07:27
What if that picture links to all of your pictures?
142
447685
2391
現在,假設這些照片,全部都與你的照片互相連結,那將會怎樣?
那時,所有這些語義資訊相互連結的
07:30
The amount of semantic interconnection and richness
143
450100
2413
資訊量將是
07:32
that comes out of that is really huge.
144
452537
1854
非常巨大的。這是非常典型的網路效應。
07:34
It's a classic network effect.
145
454415
1449
07:35
CA: Truly incredible. Congratulations.
146
455888
2024
CA:Blaise, 太難以置信了。祝賀你們!
BAA: 非常感謝各位!
關於本網站

本網站將向您介紹對學習英語有用的 YouTube 視頻。 您將看到來自世界各地的一流教師教授的英語課程。 雙擊每個視頻頁面上顯示的英文字幕,從那裡播放視頻。 字幕與視頻播放同步滾動。 如果您有任何意見或要求,請使用此聯繫表與我們聯繫。

https://forms.gle/WvT1wiN1qDtmnspy7