TED2015
Fei-Fei Li: How we're teaching computers to understand pictures
Fei-Fei Li: Kako učimo računala da razumiju slike
Filmed:
Readability: 4.5
2,702,344 views
Kada vrlo mala djevojčica pogleda sliku, ona može prepoznati osnovne elemente: "mačku", "knjigu", "stolicu". Sad, računala postaju dovoljno pametna da i ona to rade. Što je slijedeće? U ovom uzbudljivom govoru, ekspert za računalni vid Fei-Fei Li opisuje najnovija (uključujući bazu podataka od 15 milijuna slika koje je njen tim stvorio kako bi "naučio" računalo da prepoznaje slike) i glavna dostignuća.
Fei-Fei Li - Computer scientist
As Director of Stanford’s Artificial Intelligence Lab and Vision Lab, Fei-Fei Li is working to solve AI’s trickiest problems -- including image recognition, learning and language processing. Full bio
As Director of Stanford’s Artificial Intelligence Lab and Vision Lab, Fei-Fei Li is working to solve AI’s trickiest problems -- including image recognition, learning and language processing. Full bio
Double-click the English transcript below to play the video.
00:14
Let me show you something.
0
2366
3738
Dopustite da vam pokažem nešto
00:18
(Video) Girl: Okay, that's a cat
sitting in a bed.
sitting in a bed.
1
6104
4156
[Video] Djevojka: Dobro, to je mačka
koja sjedi na krevetu.
koja sjedi na krevetu.
00:22
The boy is petting the elephant.
2
10260
4040
Dječak mazi slona.
00:26
Those are people
that are going on an airplane.
that are going on an airplane.
3
14300
4354
Ovo su ljudi
koji idu u avion.
koji idu u avion.
00:30
That's a big airplane.
4
18654
2810
To je veliki avion.
Fei-Fei Li: Ovo je
trogodišnje dijete
trogodišnje dijete
00:33
Fei-Fei Li: This is
a three-year-old child
a three-year-old child
5
21464
2206
00:35
describing what she sees
in a series of photos.
in a series of photos.
6
23670
3679
koje opisuje što vidi
na ovim slikama.
na ovim slikama.
00:39
She might still have a lot
to learn about this world,
to learn about this world,
7
27349
2845
Iako ima još dosta toga
što mora naučiti o svijetu
što mora naučiti o svijetu
00:42
but she's already an expert
at one very important task:
at one very important task:
8
30194
4549
već je ekspert
u nečemu važnom:
u nečemu važnom:
00:46
to make sense of what she sees.
9
34743
2846
razumije što vidi.
00:50
Our society is more
technologically advanced than ever.
technologically advanced than ever.
10
38229
4226
Naše društvo je tehnološki
naprednije no ikada.
naprednije no ikada.
00:54
We send people to the moon,
we make phones that talk to us
we make phones that talk to us
11
42455
3629
Šaljemo ljude na mjesec,
izrađujemo telefone koji pričaju s nama
izrađujemo telefone koji pričaju s nama
00:58
or customize radio stations
that can play only music we like.
that can play only music we like.
12
46084
4946
i prilagođene radio stanice
koje puštaju samo glazbu koju volimo.
koje puštaju samo glazbu koju volimo.
01:03
Yet, our most advanced
machines and computers
machines and computers
13
51030
4055
Ipak, naši najnapredniji
uređaj i računala
uređaj i računala
01:07
still struggle at this task.
14
55085
2903
imaju poteškoća s ovim zadatkom.
01:09
So I'm here today
to give you a progress report
to give you a progress report
15
57988
3459
Ovdje sam danas
kako bi vas izvijestila
kako bi vas izvijestila
01:13
on the latest advances
in our research in computer vision,
in our research in computer vision,
16
61447
4047
o najnovijim dostignućima
u istraživanju računalnog vida,
u istraživanju računalnog vida,
01:17
one of the most frontier
and potentially revolutionary
and potentially revolutionary
17
65494
4161
jednoj od glavnih
i potencijalno revolucionarnih
i potencijalno revolucionarnih
01:21
technologies in computer science.
18
69655
3206
tehnologija računarstva.
01:24
Yes, we have prototyped cars
that can drive by themselves,
that can drive by themselves,
19
72861
4551
Imamo prototipe auta
koji se sami voze,
koji se sami voze,
01:29
but without smart vision,
they cannot really tell the difference
they cannot really tell the difference
20
77412
3853
ali bez pametnog vida,
ne mogu zapravo vidjeti razliku
ne mogu zapravo vidjeti razliku
01:33
between a crumpled paper bag
on the road, which can be run over,
on the road, which can be run over,
21
81265
3970
između zgužvane papirnate vrećice
na putu, koju mogu pregaziti,
na putu, koju mogu pregaziti,
01:37
and a rock that size,
which should be avoided.
which should be avoided.
22
85235
3340
i kamena te veličine
koji treba izbjeći.
koji treba izbjeći.
01:41
We have made fabulous megapixel cameras,
23
89415
3390
imamo odlične megapikselne kamere,
01:44
but we have not delivered
sight to the blind.
sight to the blind.
24
92805
3135
ali nismo dali
vid slijepima.
vid slijepima.
01:48
Drones can fly over massive land,
25
96420
3305
Dronovi mogu letjeti vrlo daleko
01:51
but don't have enough vision technology
26
99725
2134
ali nemaju dovoljno tehnologije vida
01:53
to help us to track
the changes of the rainforests.
the changes of the rainforests.
27
101859
3461
da nam pomognu pratiti
promjene u kišnim šumama
promjene u kišnim šumama
01:57
Security cameras are everywhere,
28
105320
2950
Sigurnosne kamere su svugdje,
02:00
but they do not alert us when a child
is drowning in a swimming pool.
is drowning in a swimming pool.
29
108270
5067
ali ne upozoravaju nas kada se
dijete utaplja u bazenu.
dijete utaplja u bazenu.
02:06
Photos and videos are becoming
an integral part of global life.
an integral part of global life.
30
114167
5595
Slike i videi postaju
integralni dio globalnog života.
integralni dio globalnog života.
02:11
They're being generated at a pace
that's far beyond what any human,
that's far beyond what any human,
31
119762
4087
Stvaraju se brzinom koja je
daleko veća od od one koji bi čovjek
daleko veća od od one koji bi čovjek
02:15
or teams of humans, could hope to view,
32
123849
2783
ili timovi ljudi željeli vidjeti,
02:18
and you and I are contributing
to that at this TED.
to that at this TED.
33
126632
3921
a vi i ja pridonosimo tome
ovdje na TED-u.
ovdje na TED-u.
02:22
Yet our most advanced software
is still struggling at understanding
is still struggling at understanding
34
130553
5232
Ipak naš najnapredniji softver
se i dalje muči oko razumjevanja
se i dalje muči oko razumjevanja
02:27
and managing this enormous content.
35
135785
3876
i upravljanja tog ogromnog sadržaja.
02:31
So in other words,
collectively as a society,
collectively as a society,
36
139661
5272
Drugim riječima,
zajedno kao društvo,
zajedno kao društvo,
poprilično smo slijepi,
02:36
we're very much blind,
37
144933
1746
jer su naši najpametniji
uređaji i dalje slijepi.
uređaji i dalje slijepi.
02:38
because our smartest
machines are still blind.
machines are still blind.
38
146679
3387
02:43
"Why is this so hard?" you may ask.
39
151526
2926
"Zašto je to tako teško?",
možda se pitate.
možda se pitate.
02:46
Cameras can take pictures like this one
40
154452
2693
Kamere mogu fotografirati slike poput ove
pretvarajući svjetlost u
dvodimenzionalne redove brojeva
dvodimenzionalne redove brojeva
02:49
by converting lights into
a two-dimensional array of numbers
a two-dimensional array of numbers
41
157145
3994
poznate kao pikseli,
02:53
known as pixels,
42
161139
1650
02:54
but these are just lifeless numbers.
43
162789
2251
ali to su samo beživotni brojevi.
02:57
They do not carry meaning in themselves.
44
165040
3111
Ne nose smisao u sebi.
03:00
Just like to hear is not
the same as to listen,
the same as to listen,
45
168151
4343
Jednako kao što slušati
ne znači isto što i čuti,
ne znači isto što i čuti,
03:04
to take pictures is not
the same as to see,
the same as to see,
46
172494
4040
fotografirati sliku nije isto
što i vidjeti,
što i vidjeti,
03:08
and by seeing,
we really mean understanding.
we really mean understanding.
47
176534
3829
a pod vidjeti
mislimo na razumijevanje.
mislimo na razumijevanje.
03:13
In fact, it took Mother Nature
540 million years of hard work
540 million years of hard work
48
181293
6177
Zapravo, prirodi je bilo potrebno
540 milijuna godina teškog posla
540 milijuna godina teškog posla
03:19
to do this task,
49
187470
1973
da to uspije,
03:21
and much of that effort
50
189443
1881
a većina tog posla
03:23
went into developing the visual
processing apparatus of our brains,
processing apparatus of our brains,
51
191324
5271
otišla je u razvijanje uređaja
za obradu vida u našem mozgu,
za obradu vida u našem mozgu,
ne u samim očima.
03:28
not the eyes themselves.
52
196595
2647
03:31
So vision begins with the eyes,
53
199242
2747
Vid započinje s očima,
03:33
but it truly takes place in the brain.
54
201989
3518
ali zapravo se sve događa u mozgu.
03:38
So for 15 years now, starting
from my Ph.D. at Caltech
from my Ph.D. at Caltech
55
206287
5060
Već 15 godina, započevši od
mog doktorata u Caltech-u
mog doktorata u Caltech-u
03:43
and then leading Stanford's Vision Lab,
56
211347
2926
i zatim vodeći Stanfordov
laboratorij za vid,
laboratorij za vid,
03:46
I've been working with my mentors,
collaborators and students
collaborators and students
57
214273
4396
radila sam s mentorima,
suradnicima i studentima
suradnicima i studentima
03:50
to teach computers to see.
58
218669
2889
kako bi naučili računala da vide.
03:54
Our research field is called
computer vision and machine learning.
computer vision and machine learning.
59
222658
3294
Naše polje se zove
računarni vid i strojno učenje.
računarni vid i strojno učenje.
03:57
It's part of the general field
of artificial intelligence.
of artificial intelligence.
60
225952
3878
Dio je većeg polja
umjetne inteligencije.
umjetne inteligencije.
04:03
So ultimately, we want to teach
the machines to see just like we do:
the machines to see just like we do:
61
231000
5493
Naposljetku, želimo naučiti
uređaje da vide kao što mi vidimo:
uređaje da vide kao što mi vidimo:
04:08
naming objects, identifying people,
inferring 3D geometry of things,
inferring 3D geometry of things,
62
236493
5387
imenovanje objekata, prepoznavanje ljudi,
razumjevanje trodimenzionalnosti objekata,
razumjevanje trodimenzionalnosti objekata,
04:13
understanding relations, emotions,
actions and intentions.
actions and intentions.
63
241880
5688
razumjevanje odnosa, emocija
akcija i namjera.
akcija i namjera.
04:19
You and I weave together entire stories
of people, places and things
of people, places and things
64
247568
6153
Vi i ja vidimo cijele priče
ljudi, mjesta i stvari
ljudi, mjesta i stvari
04:25
the moment we lay our gaze on them.
65
253721
2164
u trenutku kada ih pogledamo.
04:28
The first step towards this goal
is to teach a computer to see objects,
is to teach a computer to see objects,
66
256955
5583
Prvi korak do ovog cilja je
naučiti računala da vide objekte,
naučiti računala da vide objekte,
04:34
the building block of the visual world.
67
262538
3368
građevne jedinice vizualnog svijeta.
04:37
In its simplest terms,
imagine this teaching process
imagine this teaching process
68
265906
4434
U svom najjednostavnijem obliku,
zamislite ovaj proces učenja
zamislite ovaj proces učenja
04:42
as showing the computers
some training images
some training images
69
270340
2995
kao pokazivanje računalu
raznih prizora za trening
raznih prizora za trening
04:45
of a particular object, let's say cats,
70
273335
3321
određenog objekta, recimo mačaka,
04:48
and designing a model that learns
from these training images.
from these training images.
71
276656
4737
i dizajniranje modela koji uči
iz ovih prikaza za .
iz ovih prikaza za .
04:53
How hard can this be?
72
281393
2044
Koliko teško to može biti?
04:55
After all, a cat is just
a collection of shapes and colors,
a collection of shapes and colors,
73
283437
4052
Nakon svega, mačka je samo
skup oblika i boja,
skup oblika i boja,
04:59
and this is what we did
in the early days of object modeling.
in the early days of object modeling.
74
287489
4086
i ovo je ono što smo radili
u početcima modeliranja objekta.
u početcima modeliranja objekta.
05:03
We'd tell the computer algorithm
in a mathematical language
in a mathematical language
75
291575
3622
Napisali bi računalu algoritme
u matematičkom jeziku
u matematičkom jeziku
05:07
that a cat has a round face,
a chubby body,
a chubby body,
76
295197
3343
da mačka ima okruglo lice,
debeljuškasto tijelo,
debeljuškasto tijelo,
05:10
two pointy ears, and a long tail,
77
298540
2299
dva šiljata uha i dugačak rep,
05:12
and that looked all fine.
78
300839
1410
i da izgleda lijepo.
05:14
But what about this cat?
79
302859
2113
ali što je s ovom mačkom?
05:16
(Laughter)
80
304972
1091
(Smijeh)
05:18
It's all curled up.
81
306063
1626
Sva je izvijena.
05:19
Now you have to add another shape
and viewpoint to the object model.
and viewpoint to the object model.
82
307689
4719
Sad morate dodati drugi oblik
i pogled modelnom objektu.
i pogled modelnom objektu.
05:24
But what if cats are hidden?
83
312408
1715
Što ako su mačke skrivene?
05:27
What about these silly cats?
84
315143
2219
Što je sa smiješnim mačkama?
05:31
Now you get my point.
85
319112
2417
Sad vidite što želim reći.
05:33
Even something as simple
as a household pet
as a household pet
86
321529
3367
Čak i nešto jednostavno
poput kućnog ljubimca
poput kućnog ljubimca
05:36
can present an infinite number
of variations to the object model,
of variations to the object model,
87
324896
4504
može imati beskonačan broj
varijacija modelnog objekta,
varijacija modelnog objekta,
05:41
and that's just one object.
88
329400
2233
i to je samo jedan objekt.
05:44
So about eight years ago,
89
332573
2492
Prije osam godina,
05:47
a very simple and profound observation
changed my thinking.
changed my thinking.
90
335065
5030
vrlo jednostavno i duboko zapažanje
promjenilo mi je razmišljanje.
promjenilo mi je razmišljanje.
05:53
No one tells a child how to see,
91
341425
2685
Nitko ne govori djetetu kako da vidi,
05:56
especially in the early years.
92
344110
2261
posebno u ranijim godinama.
05:58
They learn this through
real-world experiences and examples.
real-world experiences and examples.
93
346371
5000
Oni to uče kroz
iskustvo i primjere iz stvarnog svijeta.
iskustvo i primjere iz stvarnog svijeta.
06:03
If you consider a child's eyes
94
351371
2740
Ako smatrate dječje oči
06:06
as a pair of biological cameras,
95
354111
2554
parom bioloških kamera,
06:08
they take one picture
about every 200 milliseconds,
about every 200 milliseconds,
96
356665
4180
one fotografiraju
svakih 200 milisekundi,
svakih 200 milisekundi,
06:12
the average time an eye movement is made.
97
360845
3134
prosječno vrijeme koliko je potrebno
za pokret oka.
za pokret oka.
06:15
So by age three, a child would have seen
hundreds of millions of pictures
hundreds of millions of pictures
98
363979
5550
Do svoje treće godine, dijete bi vidjelo
stotine milijuna slika
stotine milijuna slika
06:21
of the real world.
99
369529
1834
stvarnog svijeta.
06:23
That's a lot of training examples.
100
371363
2280
To je puno primjera za vježbu.
06:26
So instead of focusing solely
on better and better algorithms,
on better and better algorithms,
101
374383
5989
Umjesto fokusiranja samo na
sve bolje i bolje algoritme,
sve bolje i bolje algoritme,
06:32
my insight was to give the algorithms
the kind of training data
the kind of training data
102
380372
5272
mislila sam dati algoritmima
nekakakve podatke za vježbu
nekakakve podatke za vježbu
06:37
that a child was given through experiences
103
385644
3319
koje je dijete dobijalo kroz iskustva
06:40
in both quantity and quality.
104
388963
3878
i to kvantitativno i kvalitativno.
06:44
Once we know this,
105
392841
1858
Jednom kada znamo ovo,
06:46
we knew we needed to collect a data set
106
394699
2971
znali smo da moramo skupiti skup podataka
06:49
that has far more images
than we have ever had before,
than we have ever had before,
107
397670
4459
koji ima puno više prikaza
no što smo mi imali ikad prije,
no što smo mi imali ikad prije,
06:54
perhaps thousands of times more,
108
402129
2577
možda i tisuću puta više,
06:56
and together with Professor
Kai Li at Princeton University,
Kai Li at Princeton University,
109
404706
4111
i zajedno s profesorom
Kai Li na sveučilištu Princeton,
Kai Li na sveučilištu Princeton,
07:00
we launched the ImageNet project in 2007.
110
408817
4752
2007. lansirali smo ImageNet projekt.
07:05
Luckily, we didn't have to mount
a camera on our head
a camera on our head
111
413569
3838
Sva sreća nismo morali montirati
kamere na naše glave
kamere na naše glave
07:09
and wait for many years.
112
417407
1764
i čekati godinama.
Otišli smo na Internet,
07:11
We went to the Internet,
113
419171
1463
07:12
the biggest treasure trove of pictures
that humans have ever created.
that humans have ever created.
114
420634
4436
najveću riznicu slika koju je
čovječanstvo stvorilo.
čovječanstvo stvorilo.
07:17
We downloaded nearly a billion images
115
425070
3041
skinuli smo skoro
milijardu slika i
milijardu slika i
07:20
and used crowdsourcing technology
like the Amazon Mechanical Turk platform
like the Amazon Mechanical Turk platform
116
428111
5880
koristili crowdsourcing tehnologiju
poput platforme Amazon Mechanical Turk
poput platforme Amazon Mechanical Turk
07:25
to help us to label these images.
117
433991
2339
da označimo te prikaze.
07:28
At its peak, ImageNet was one of
the biggest employers
the biggest employers
118
436330
4900
Kako je raslo, ImageNet je bio jedan od
najvećih poslodavaca
najvećih poslodavaca
07:33
of the Amazon Mechanical Turk workers:
119
441230
2996
radnika Amazon Mechanical Turk-a:
07:36
together, almost 50,000 workers
120
444226
3854
zajedno, skoro 50.000 radnika
07:40
from 167 countries around the world
121
448080
4040
iz 167 država svijeta
07:44
helped us to clean, sort and label
122
452120
3947
pomoglo nam je da očistimo,
sortiramo i označimo
sortiramo i označimo
07:48
nearly a billion candidate images.
123
456067
3575
skoro milijardu korisnih prikaza.
07:52
That was how much effort it took
124
460612
2653
Toliko truda je trebalo
07:55
to capture even a fraction
of the imagery
of the imagery
125
463265
3900
da se uhvati dio prikaza
07:59
a child's mind takes in
in the early developmental years.
in the early developmental years.
126
467165
4171
koje djetetov um uhvati
u ranim godinama razvoja.
u ranim godinama razvoja.
08:04
In hindsight, this idea of using big data
127
472148
3902
Na očigled, ova ideja
korištenja mnogo podataka
korištenja mnogo podataka
08:08
to train computer algorithms
may seem obvious now,
may seem obvious now,
128
476050
4550
da se istreniraju računalni algoritmi
se možda sada čini očiglednim,
se možda sada čini očiglednim,
08:12
but back in 2007, it was not so obvious.
129
480600
4110
ali 2007., nije bilo tako očigledno.
08:16
We were fairly alone on this journey
for quite a while.
for quite a while.
130
484710
3878
Prilično dugo bili smo
poprilično sami na tom putu.
poprilično sami na tom putu.
08:20
Some very friendly colleagues advised me
to do something more useful for my tenure,
to do something more useful for my tenure,
131
488588
5003
Neke prijateljski nastrojene kolege su
me savjetovale da radim nešto korisnije,
me savjetovale da radim nešto korisnije,
08:25
and we were constantly struggling
for research funding.
for research funding.
132
493591
4342
i cijelo vrijeme smo se borili
za financiranje istraživanja.
za financiranje istraživanja.
08:29
Once, I even joked to my graduate students
133
497933
2485
Jednom, sam se čak našalila sa studentima
08:32
that I would just reopen
my dry cleaner's shop to fund ImageNet.
my dry cleaner's shop to fund ImageNet.
134
500418
4063
da ću ponovno otvoriti kemijsku čistionicu
kako bih mogla financirati ImageNet.
kako bih mogla financirati ImageNet.
08:36
After all, that's how I funded
my college years.
my college years.
135
504481
4761
Naposljetku, tako sam financirala
svoj studij.
svoj studij.
08:41
So we carried on.
136
509242
1856
Nastavili smo dalje.
08:43
In 2009, the ImageNet project delivered
137
511098
3715
2009. ImageNet je dosegao
08:46
a database of 15 million images
138
514813
4042
bazu podataka od 15 milijuna prikaza
08:50
across 22,000 classes
of objects and things
of objects and things
139
518855
4805
preko 22.000 klasa
objekata i stvari
objekata i stvari
08:55
organized by everyday English words.
140
523660
3320
organiziranih u svakodnevne
engleske riječi.
engleske riječi.
08:58
In both quantity and quality,
141
526980
2926
I po kvantiteti i po kvaliteti
09:01
this was an unprecedented scale.
142
529906
2972
ovo je dosad nedostignuta skala.
09:04
As an example, in the case of cats,
143
532878
3461
Kao primjer, u slučaju mačaka,
09:08
we have more than 62,000 cats
144
536339
2809
imamo više od 62.000 mačaka
09:11
of all kinds of looks and poses
145
539148
4110
u svim oblicima i pozama,
09:15
and across all species
of domestic and wild cats.
of domestic and wild cats.
146
543258
5223
i različitih vrsta
domaćih i divljih mačaka.
domaćih i divljih mačaka.
09:20
We were thrilled
to have put together ImageNet,
to have put together ImageNet,
147
548481
3344
Bili smo oduševljeni
što smo sastavili ImageNet,
što smo sastavili ImageNet,
09:23
and we wanted the whole research world
to benefit from it,
to benefit from it,
148
551825
3738
i htjeli smo da cijeli znanstveni svijet
ima koristi od njega,
ima koristi od njega,
09:27
so in the TED fashion,
we opened up the entire data set
we opened up the entire data set
149
555563
4041
tako da smo po modi TED-a
otvorili cijeli skup podataka
otvorili cijeli skup podataka
09:31
to the worldwide
research community for free.
research community for free.
150
559604
3592
svim istraživačkim zajednicama, besplatno.
(Pljesak)
09:36
(Applause)
151
564636
4000
09:41
Now that we have the data
to nourish our computer brain,
to nourish our computer brain,
152
569416
4538
Sad kad imamo podatke
da opskrbimo mozgove naših računala,
da opskrbimo mozgove naših računala,
09:45
we're ready to come back
to the algorithms themselves.
to the algorithms themselves.
153
573954
3737
spremni smo vratiti se na
same algoritme.
same algoritme.
09:49
As it turned out, the wealth
of information provided by ImageNet
of information provided by ImageNet
154
577691
5178
Ispalo je kako je bogatstvo informacija
s ImageNet-a
s ImageNet-a
09:54
was a perfect match to a particular class
of machine learning algorithms
of machine learning algorithms
155
582869
4806
savršeno za određene vrste
algoritama za strojno učenje
algoritama za strojno učenje
09:59
called convolutional neural network,
156
587675
2415
koji se zovu konvolucijske neuronske mreže
10:02
pioneered by Kunihiko Fukushima,
Geoff Hinton, and Yann LeCun
Geoff Hinton, and Yann LeCun
157
590090
5248
osmišljene od strane Kunihiko Fukushime,
Geoff Hintona i Yann LeCuna
Geoff Hintona i Yann LeCuna
10:07
back in the 1970s and '80s.
158
595338
3645
davnih 1970-ih i 1980-ih.
10:10
Just like the brain consists
of billions of highly connected neurons,
of billions of highly connected neurons,
159
598983
5619
Upravo kako se mozak sastoji
od milijardu vrlo povezanih neurona,
od milijardu vrlo povezanih neurona,
10:16
a basic operating unit in a neural network
160
604602
3854
osnovna operacijska jedinica
neuronskih mreža
neuronskih mreža
10:20
is a neuron-like node.
161
608456
2415
jest čvor sličan neuronu.
10:22
It takes input from other nodes
162
610871
2554
Prima podatke od drugih čvorova
10:25
and sends output to others.
163
613425
2718
i šalje ih drugima.
Ove stotine tisuća ili
čak milijuni čvorova
čak milijuni čvorova
10:28
Moreover, these hundreds of thousands
or even millions of nodes
or even millions of nodes
164
616143
4713
10:32
are organized in hierarchical layers,
165
620856
3227
su organizirani po
hijerarhijskim slojevima
hijerarhijskim slojevima
10:36
also similar to the brain.
166
624083
2554
sličnim onima u mozgu.
10:38
In a typical neural network we use
to train our object recognition model,
to train our object recognition model,
167
626637
4783
U tipičnoj neuralnoj mreži koju koristimo
u učenju prepoznavanja modela,
u učenju prepoznavanja modela,
10:43
it has 24 million nodes,
168
631420
3181
ima 24 milijuna čvorova,
10:46
140 million parameters,
169
634601
3297
140 milijuna parametara,
10:49
and 15 billion connections.
170
637898
2763
i 15 milijardi veza.
10:52
That's an enormous model.
171
640661
2415
To je ogroman model.
10:55
Powered by the massive data from ImageNet
172
643076
3901
Upogonjen je s
mnoštvom podataka s ImageNet-a
mnoštvom podataka s ImageNet-a
10:58
and the modern CPUs and GPUs
to train such a humongous model,
to train such a humongous model,
173
646977
5433
te modernih CPJ-a i GPJ-a
kako bi istrenirao ove ogrome modele,
kako bi istrenirao ove ogrome modele,
11:04
the convolutional neural network
174
652410
2369
skupna neuronska mreža
11:06
blossomed in a way that no one expected.
175
654779
3436
je procvala na način
koji nitko nije očekivao.
koji nitko nije očekivao.
11:10
It became the winning architecture
176
658215
2508
Postala je ključna struktura
11:12
to generate exciting new results
in object recognition.
in object recognition.
177
660723
5340
koja je dovodila do novih uzbudljivih
rezultata u prepoznavanju objekata.
rezultata u prepoznavanju objekata.
11:18
This is a computer telling us
178
666063
2810
Ovo je računalo koje nam govori
11:20
this picture contains a cat
179
668873
2300
da je na slici mačka
11:23
and where the cat is.
180
671173
1903
i gdje je mačka.
11:25
Of course there are more things than cats,
181
673076
2112
Naravno ne radi se samo o mački,
11:27
so here's a computer algorithm telling us
182
675188
2438
ovdje nam računalni algoritam govori
11:29
the picture contains
a boy and a teddy bear;
a boy and a teddy bear;
183
677626
3274
da slika sadrži
dječaka i medvjedića;
dječaka i medvjedića;
11:32
a dog, a person, and a small kite
in the background;
in the background;
184
680900
4366
psa, osobu i malog zmaja
u pozadini;
u pozadini;
11:37
or a picture of very busy things
185
685266
3135
ili slika vrlo zbrkanih stvari
11:40
like a man, a skateboard,
railings, a lampost, and so on.
railings, a lampost, and so on.
186
688401
4644
poput čovjeka, skateboarda,
ograde, lampe itd.
ograde, lampe itd.
11:45
Sometimes, when the computer
is not so confident about what it sees,
is not so confident about what it sees,
187
693045
5293
Ponekad kada računalo
nije sigurno što vidi,
nije sigurno što vidi,
11:51
we have taught it to be smart enough
188
699498
2276
moramo ga naučiti da bude dovoljno pametno
11:53
to give us a safe answer
instead of committing too much,
instead of committing too much,
189
701774
3878
da nam pruži siguran odgovor,
11:57
just like we would do,
190
705652
2811
kao što bismo mi odgovorili,
12:00
but other times our computer algorithm
is remarkable at telling us
is remarkable at telling us
191
708463
4666
ali u drugim slučajevima računalni
alogoritam nam besprijekorno kaže
alogoritam nam besprijekorno kaže
12:05
what exactly the objects are,
192
713129
2253
što su točno ti objekti,
12:07
like the make, model, year of the cars.
193
715382
3436
poput materijala, modela, godine auta.
12:10
We applied this algorithm to millions
of Google Street View images
of Google Street View images
194
718818
5386
Primjenili smo ovaj algoritam na milijune
Google Street View prikaza
Google Street View prikaza
12:16
across hundreds of American cities,
195
724204
3135
u stotinama američkih gradova,
12:19
and we have learned something
really interesting:
really interesting:
196
727339
2926
i spoznali smo nešto vrlo zanimljivo:
12:22
first, it confirmed our common wisdom
197
730265
3320
prvo, potvrdilo se staro pravilo
12:25
that car prices correlate very well
198
733585
3290
da cijene auta dobro koreliraju
12:28
with household incomes.
199
736875
2345
s kućnim primanjima.
12:31
But surprisingly, car prices
also correlate well
also correlate well
200
739220
4527
Ali isto tako cijene auta
koreliraju također sa
koreliraju također sa
12:35
with crime rates in cities,
201
743747
2300
stopom kriminala u gradovima,
12:39
or voting patterns by zip codes.
202
747007
3963
ili načina glasanja po poštanskom broju.
12:44
So wait a minute. Is that it?
203
752060
2206
Čekajte.
Je li to, to?
Je li to, to?
12:46
Has the computer already matched
or even surpassed human capabilities?
or even surpassed human capabilities?
204
754266
5153
Je li nas računalo već sustigao
ili čak prestiglo u našim sposobnostima?
ili čak prestiglo u našim sposobnostima?
12:51
Not so fast.
205
759419
2138
Ne tako brzo.
12:53
So far, we have just taught
the computer to see objects.
the computer to see objects.
206
761557
4923
Zasad smo samo naučili
računalo da vidi objekte.
računalo da vidi objekte.
12:58
This is like a small child
learning to utter a few nouns.
learning to utter a few nouns.
207
766480
4644
To je kao da malo dijete
učite reći nekoliko imenica.
učite reći nekoliko imenica.
13:03
It's an incredible accomplishment,
208
771124
2670
To je ogromno postignuće,
ali je to tek prvi korak.
13:05
but it's only the first step.
209
773794
2460
13:08
Soon, another developmental
milestone will be hit,
milestone will be hit,
210
776254
3762
Uskoro će drugo razvojno postignuće
biti dosegnuto,
biti dosegnuto,
13:12
and children begin
to communicate in sentences.
to communicate in sentences.
211
780016
3461
i djeca počinju komunicirati u rečenicama.
13:15
So instead of saying
this is a cat in the picture,
this is a cat in the picture,
212
783477
4224
Stoga umjesto govorenja
kako je mačka na slici,
kako je mačka na slici,
13:19
you already heard the little girl
telling us this is a cat lying on a bed.
telling us this is a cat lying on a bed.
213
787701
5202
već ste čuli malu djevojčicu
koja govori da mačka leži na krevetu.
koja govori da mačka leži na krevetu.
13:24
So to teach a computer
to see a picture and generate sentences,
to see a picture and generate sentences,
214
792903
5595
Kako bi naučili računalo
da vidi sliku i stvori rečenice,
da vidi sliku i stvori rečenice,
13:30
the marriage between big data
and machine learning algorithm
and machine learning algorithm
215
798498
3948
brak između velikih podataka
i algoritama strojnog učenja
i algoritama strojnog učenja
13:34
has to take another step.
216
802446
2275
mora ići korak dalje.
13:36
Now, the computer has to learn
from both pictures
from both pictures
217
804721
4156
Računalo mora naučiti
učiti i iz slika
učiti i iz slika
13:40
as well as natural language sentences
218
808877
2856
i iz prirodnih jezičnih rečenica
13:43
generated by humans.
219
811733
3322
stvorenih od strane ljudi.
13:47
Just like the brain integrates
vision and language,
vision and language,
220
815055
3853
Upravo kako mozak integrira
vid i jezik,
vid i jezik,
13:50
we developed a model
that connects parts of visual things
that connects parts of visual things
221
818908
5201
razvili smo model
koji spaja vidljive dijelove
koji spaja vidljive dijelove
13:56
like visual snippets
222
824109
1904
poput vidnih komada
13:58
with words and phrases in sentences.
223
826013
4203
s riječima i frazama u rečenicama.
14:02
About four months ago,
224
830216
2763
Otprilike prije četiri mjeseca,
14:04
we finally tied all this together
225
832979
2647
konačno smo uspjelo sve povezati
14:07
and produced one of the first
computer vision models
computer vision models
226
835626
3784
i proizveli smo jedan od prvih
modela računalnog vida
modela računalnog vida
14:11
that is capable of generating
a human-like sentence
a human-like sentence
227
839410
3994
koji je sposoban stvoriti
rečenicu sličnu ljudskoj
rečenicu sličnu ljudskoj
14:15
when it sees a picture for the first time.
228
843404
3506
kada vidi sliku po prvi puta.
14:18
Now, I'm ready to show you
what the computer says
what the computer says
229
846910
4644
Pokazat ću vam
što računalo kaže
što računalo kaže
14:23
when it sees the picture
230
851554
1975
kada vidi slike
14:25
that the little girl saw
at the beginning of this talk.
at the beginning of this talk.
231
853529
3830
koje je mala djevojčica vidjela
na početku govora.
na početku govora.
14:31
(Video) Computer: A man is standing
next to an elephant.
next to an elephant.
232
859519
3344
(Video) Računalo: Čovjek stoji
pored slona.
pored slona.
14:36
A large airplane sitting on top
of an airport runway.
of an airport runway.
233
864393
3634
Veliki avion sjedi na vrhu
avionske piste.
avionske piste.
14:41
FFL: Of course, we're still working hard
to improve our algorithms,
to improve our algorithms,
234
869057
4212
FFL: Naravno, i dalje se trudimo
unaprijediti naše algoritme,
unaprijediti naše algoritme,
14:45
and it still has a lot to learn.
235
873269
2596
i još puno toga mora naučiti.
(Pljesak)
14:47
(Applause)
236
875865
2291
14:51
And the computer still makes mistakes.
237
879556
3321
I računalo i dalje pravi greške.
(Video) Računalo: Mačka leži
na krevetu u deci.
na krevetu u deci.
14:54
(Video) Computer: A cat lying
on a bed in a blanket.
on a bed in a blanket.
238
882877
3391
FFL: Naravno, kada vidi
previše mačaka,
previše mačaka,
14:58
FFL: So of course, when it sees
too many cats,
too many cats,
239
886268
2553
15:00
it thinks everything
might look like a cat.
might look like a cat.
240
888821
2926
misli da bi sve moglo
izgledati kao mačka.
izgledati kao mačka.
15:05
(Video) Computer: A young boy
is holding a baseball bat.
is holding a baseball bat.
241
893317
2864
(Video) Računalo: Dječak
drži bejzbolsku palicu.
drži bejzbolsku palicu.
15:08
(Laughter)
242
896181
1765
(Smijeh)
15:09
FFL: Or, if it hasn't seen a toothbrush,
it confuses it with a baseball bat.
it confuses it with a baseball bat.
243
897946
4583
FFL: Ili, ako nije vidio četkicu za zube,
pomiješat će je s bejzbolskom palicom.
pomiješat će je s bejzbolskom palicom.
15:15
(Video) Computer: A man riding a horse
down a street next to a building.
down a street next to a building.
244
903309
3434
(Video) Računalo: Čovjek jaše konja
niz ulicu pored zgrade.
niz ulicu pored zgrade.
15:18
(Laughter)
245
906743
2023
(Smijeh)
15:20
FFL: We haven't taught Art 101
to the computers.
to the computers.
246
908766
3552
FFL: Nismo računalo naučili
neke osnove umjetnosti.
neke osnove umjetnosti.
15:25
(Video) Computer: A zebra standing
in a field of grass.
in a field of grass.
247
913768
2884
(Video) Računalo: Zebra stoji
u polju trave.
u polju trave.
15:28
FFL: And it hasn't learned to appreciate
the stunning beauty of nature
the stunning beauty of nature
248
916652
3367
FFL: I nije naučio diviti se
prekrasnoj ljepoti prirode
prekrasnoj ljepoti prirode
15:32
like you and I do.
249
920019
2438
kao vi i ja.
15:34
So it has been a long journey.
250
922457
2832
Bilo je to dugo putovanje.
15:37
To get from age zero to three was hard.
251
925289
4226
Od rođenja do treće godine je bilo teško.
15:41
The real challenge is to go
from three to 13 and far beyond.
from three to 13 and far beyond.
252
929515
5596
Pravi izazov je doći
od treće do trinaeste godine, i dalje.
od treće do trinaeste godine, i dalje.
15:47
Let me remind you with this picture
of the boy and the cake again.
of the boy and the cake again.
253
935111
4365
Podsjetit ću vas s opet s ovom
slikom dječaka i kolača.
slikom dječaka i kolača.
15:51
So far, we have taught
the computer to see objects
the computer to see objects
254
939476
4064
Dosad smo naučili računalo
da vidi objekte
da vidi objekte
15:55
or even tell us a simple story
when seeing a picture.
when seeing a picture.
255
943540
4458
ili čak nam kaže jednostavnu priču
onoga što je na slici.
onoga što je na slici.
15:59
(Video) Computer: A person sitting
at a table with a cake.
at a table with a cake.
256
947998
3576
(Video) Računalo: Osoba sjedi
za stolom s kolačem.
za stolom s kolačem.
FFL: Ali postoji puno više
na ovoj slici
na ovoj slici
16:03
FFL: But there's so much more
to this picture
to this picture
257
951574
2630
nego samo osoba i kolač.
16:06
than just a person and a cake.
258
954204
2270
Što računalo ne vidi jest
da je to poseban talijanski kolač
da je to poseban talijanski kolač
16:08
What the computer doesn't see
is that this is a special Italian cake
is that this is a special Italian cake
259
956474
4467
16:12
that's only served during Easter time.
260
960941
3217
koji se jedino servira za vrijeme Uskrsa.
16:16
The boy is wearing his favorite t-shirt
261
964158
3205
Dječak nosi svoju omiljenu majicu
koju je dobio od oca
nakon putovanja u Sidney,
nakon putovanja u Sidney,
16:19
given to him as a gift by his father
after a trip to Sydney,
after a trip to Sydney,
262
967363
3970
16:23
and you and I can all tell how happy he is
263
971333
3808
i vi i ja možemo reći da je jako stretan
16:27
and what's exactly on his mind
at that moment.
at that moment.
264
975141
3203
i što je na njegovom umu
u ovom trenu.
u ovom trenu.
16:31
This is my son Leo.
265
979214
3125
To je moj sin Leo.
16:34
On my quest for visual intelligence,
266
982339
2624
Na mom pohodu na vidnu inteligenciju,
razmišljam o Leu konstantno
16:36
I think of Leo constantly
267
984963
2391
16:39
and the future world he will live in.
268
987354
2903
i budućnosti u kojoj će živjeti.
16:42
When machines can see,
269
990257
2021
Kada uređaji vide,
16:44
doctors and nurses will have
extra pairs of tireless eyes
extra pairs of tireless eyes
270
992278
4712
doktori i sestre će imati
dodatan par neumornih očiju
dodatan par neumornih očiju
16:48
to help them to diagnose
and take care of patients.
and take care of patients.
271
996990
4092
koje im pomažu dijagnosticirati
i pobrinuti se za pacijenta.
i pobrinuti se za pacijenta.
16:53
Cars will run smarter
and safer on the road.
and safer on the road.
272
1001082
4383
Auti će voziti pametnije
i sigurnije na putu.
i sigurnije na putu.
16:57
Robots, not just humans,
273
1005465
2694
Roboti, ne samo ljudi,
17:00
will help us to brave the disaster zones
to save the trapped and wounded.
to save the trapped and wounded.
274
1008159
4849
će pomoći u opasnim situacijama
kako bi spasili zatočene i ozljeđene.
kako bi spasili zatočene i ozljeđene.
Otkrit ćemo nove vrste,
bolje materijale,
bolje materijale,
17:05
We will discover new species,
better materials,
better materials,
275
1013798
3796
17:09
and explore unseen frontiers
with the help of the machines.
with the help of the machines.
276
1017594
4509
i istražiti neviđene granice
uz pomoć uređaja.
uz pomoć uređaja.
17:15
Little by little, we're giving sight
to the machines.
to the machines.
277
1023113
4167
Malo po malo, dajemo
vid uređajima.
vid uređajima.
17:19
First, we teach them to see.
278
1027280
2798
Prvo, smo ih naučili da vide.
17:22
Then, they help us to see better.
279
1030078
2763
Onda nam oni pomažu vidjeti bolje.
17:24
For the first time, human eyes
won't be the only ones
won't be the only ones
280
1032841
4165
Po prvi put, ljudsko oko
neće biti jedino
neće biti jedino
17:29
pondering and exploring our world.
281
1037006
2934
koje gleda i istražuje svijet.
17:31
We will not only use the machines
for their intelligence,
for their intelligence,
282
1039940
3460
Nećemo koristiti uređaje
zbog njihove inteligencije,
zbog njihove inteligencije,
17:35
we will also collaborate with them
in ways that we cannot even imagine.
in ways that we cannot even imagine.
283
1043400
6179
surađivat ćemo s njima
na načine koje ne možemo zamisliti.
na načine koje ne možemo zamisliti.
17:41
This is my quest:
284
1049579
2161
Ovo je moj pothvat:
17:43
to give computers visual intelligence
285
1051740
2712
dati računalima vidnu inteligenciju
17:46
and to create a better future
for Leo and for the world.
for Leo and for the world.
286
1054452
5131
i stvoriti bolje sutra
za Lea i za svijet.
za Lea i za svijet.
17:51
Thank you.
287
1059583
1811
Hvala vam.
(Pljesak)
17:53
(Applause)
288
1061394
3785
ABOUT THE SPEAKER
Fei-Fei Li - Computer scientistAs Director of Stanford’s Artificial Intelligence Lab and Vision Lab, Fei-Fei Li is working to solve AI’s trickiest problems -- including image recognition, learning and language processing.
Why you should listen
Using algorithms built on machine learning methods such as neural network models, the Stanford Artificial Intelligence Lab led by Fei-Fei Li has created software capable of recognizing scenes in still photographs -- and accurately describe them using natural language.
Li’s work with neural networks and computer vision (with Stanford’s Vision Lab) marks a significant step forward for AI research, and could lead to applications ranging from more intuitive image searches to robots able to make autonomous decisions in unfamiliar situations.
Fei-Fei was honored as one of Foreign Policy's 2015 Global Thinkers.
Fei-Fei Li | Speaker | TED.com