A camera that can see around corners | David Lindell

Transcriber: Ivana Korom Reviewer: Krystian Aparta

Prevodilac: Filip Eskić Lektor: Ivana Korom

In the future,

U budućnosti,

self-driving cars will be safer and more reliable than humans.

samovozeći automobili će biti bezbedniji i pouzdaniji od ljudi.

Ali, da bi se ovo desilo,

But for this to happen,

we need technologies that allow cars to respond

potrebne su nam tehnologije koje vozilima omogućavaju da reaguju

faster than humans,

brže od ljudi,

we need algorithms that can drive better than humans

potrebni su nam algoritmi koji mogu da voze bolje od ljudi

and we need cameras that can see more than humans can see.

i potrebne su nam kamere koje mogu videti više nego što čovek može da vidi.

For example, imagine a self-driving car is about to make a blind turn,

Na primer, zamislite da samovozeći auto želi da izvrši slepo skretanje,

and there's an oncoming car

a u susret mu nailazi auto

or perhaps there's a child about to run into the street.

ili će možda dete istrčati na ulicu.

Fortunately, our future car will have this superpower,

Na svu sreću, naš budući automobil će imati ovu supermoć,

a camera that can see around corners to detect these potential hazards.

kameru koja će moći da vidi oko uglova da bi otkrila ovakve potencijalne nesreće.

For the past few years as a PhD student

Nekoliko proteklih godina kao student na doktorskim studijama

in the Stanford Computational Imaging Lab,

u Stenfordovoj računarskoj laboratoriji,

I've been working on a camera that can do just this --

radim na kameri koja bi mogla da uradi to -

a camera that can image objects hidden around corners

kamera koja može da pronađe objekte koji su sakriveni u uglovima

or blocked from direct line of sight.

ili blokirani od direktne linije vidnog polja.

So let me give you an example of what our camera can see.

Dopustite da vam dam primer šta naša kamera može videti.

This is an outdoor experiment we conducted

Ovo je spoljašnji eksperiment koji smo sproveli

where our camera system is scanning the side of this building with a laser,

gde naš sistem kamere skenira ivicu zgrade laserom,

and the scene that we want to capture

i scena koju želimo da zabeležimo

is hidden around the corner behind this curtain.

je skrivena iza ugla, iza zavese.

So our camera system can't actually see it directly.

Naš sistem je ne može videti direktno.

And yet, somehow,

A ipak nekako,

our camera can still capture the 3D geometry of this scene.

naša kamera može da napravi 3D geometriju ove scene.

So how do we do this?

Kako to izvodimo?

The magic happens here in this camera system.

Magija nastaje u ovom sistemu kamere.

You can think of this as a type of high-speed camera.

Možete gledati na ovo kao na veoma brzu kameru.

Not one that operates at 1,000 frames per second,

Ne onu koja radi sa 1000 slika u sekundi,

or even a million frames per second,

ili čak milion slika u sekundi,

but a trillion frames per second.

već sa bilion slika u sekundi.

So fast that it can actually capture the movement of light itself.

Toliko je brza da zapravo može da slika kretanje same svetlosti.

And to give you an example of just how fast light travels,

A da bih vam dao primer koliko brzo svetlost putuje,

let's compare it to the speed of a fast-running comic book superhero

uporedimo to sa brzinom brzotrčećeg superheroja iz stripa

who can move at up to three times the speed of sound.

koji se može kretati tri puta većom brzinom od brzine zvuka.

It takes a pulse of light about 3.3 billionths of a second,

Pulsu svetlosti je potrebno oko 3.3 milijarditog dela sekunde

or 3.3 nanoseconds,

ili 3.3 nanosekunde,

to travel the distance of a meter.

da pređe udaljenost od jednog metra.

Well, in that same time,

Pa, za to isto vreme,

our superhero has moved less than the width of a human hair.

naš superheroj se pomerio manje od širine ljudske dlake.

That's pretty fast.

To je prilično brzo.

But actually, we need to image much faster

Ali, zapravo, potrebno nam je da stvaramo sliku mnogo brže

if we want to capture light moving at subcentimeter scales.

ako želimo da uhvatimo da se svetlost kreće subcentimetarskom razmerom.

So our camera system can capture photons

Dakle, naša kamera može da slika fotone

at time frames of just 50 trillionths of a second,

u vremenskim okvirima od samo 50 bilionitih delova sekunde,

or 50 picoseconds.

ili 50 pikosekundi.

So we take this ultra-high-speed camera

Dakle, uzimamo ovu ultra visoko brzu kameru

and we pair it with a laser that sends out short pulses of light.

i uparujemo je sa laserom koji šalje kratke impulse svetlosti.

Each pulse travels to this visible wall

Svaki puls putuje do ovog vidljivog zida

and some light scatters back to our camera,

i neka svetla se rasipaju nazad ka našoj kameri,

but we also use the wall to scatter light around the corner

ali mi takođe koristimo zid i da razbacamo svetlost iza ugla

to the hidden object and back.

ka skrivenom objektu i nazad.

We repeat this measurement many times

Ovaj postupak ponavljamo mnogo puta

to capture the arrival times of many photons

da bismo zabeležili pristizanje mnogo fotona

from different locations on the wall.

sa različitih lokacija zida.

And after we capture these measurements, we can create

Nakon što izvršimo merenja, možemo da napravimo

a trillion-frame-per-second video of the wall.

video od bilion slika po sekundi.

While this wall may look ordinary to our own eyes,

Iako ovaj zid može našim očima da izgleda obično,

at a trillion frames per second, we can see something truly incredible.

sa bilion slika po sekundi možemo videti nešto zaista neverovatno.

We can actually see waves of light scattered back from the hidden scene

Možemo zapravo videti svetlosne talase razbacane sa skrivene scene

and splashing against the wall.

kako se sudaraju sa zidom.

And each of these waves carries information

I svaki od ovih talasa nosi informaciju

about the hidden object that sent it.

o skrivenom objektu koji ga je poslao.

So we can take these measurements

Dakle, možemo uzeti ove mere

and pass them into a reconstruction algorithm

i preneti ih u algoritam za rekonstrukciju

to then recover the 3D geometry of this hidden scene.

da bismo zatim stvorili 3D geometriju skrivene scene.

Now I want to show you one more example of an indoor scene that we captured,

Želim da vam pokažem još jedan primer scene iz unutrašnjosti koju smo snimili,

this time with a variety of different hidden objects.

ovoga puta sa različitim skrivenim objektima.

And these objects have different appearances,

Ovi objekti drugačije izgledaju,

so they reflect light differently.

tako da drugačije odbijaju svetlost.

For example, this glossy dragon statue reflects light differently

Na primer, ova sjajna statua zmaja drugačije odbija svetlost

than the mirror disco ball

nego disko kugla

or the white discus thrower statue.

ili bela statua bacača diska.

And we can actually see the differences in the reflected light

I mi, zapravo, možemo videti razlike u odbijenom svetlu

by visualizing it as this 3D volume,

posmatrajući ga kao ovu 3D zapreminu,

where we've just taken the video frames and stacked them together.

gde smo samo uzeli slike iz videa i spakovali ih zajedno.

And time here is represented as the depth dimension of this cube.

Vreme je ovde predstavljeno kao dubina ove kocke.

These bright dots that you see are reflections of light

Ove sjajne tačke koje vidite su odrazi svetlosti,

from each of the mirrored facets of the disco ball,

od svakog aspekta disko kugle,

scattering against the wall over time.

šireći se po zidu tokom vremena.

The bright streaks of light that you see arriving soonest in time

Svetli tragovi svetlosti koje vidite da najbrže stižu u vremenu

are from the glossy dragon statue that's closest to the wall,

pripadaju sjajnoj zmajevoj statui koja je najbliža zidu,

and the other streaks of light come from reflections of light from the bookcase

a drugi tragovi svetlosti dolaze od odraza svetlosti od police za knjige

and from the statue.

i od statue.

Now, we can also visualize these measurements frame by frame,

Možemo takođe vizuelizovati ova merenja sliku po sliku,

as a video,

kao video,

to directly see the scattered light.

da bismo direktno videli rašireno svetlo.

And again, here we see, first, reflections of light from the dragon,

Ponovo, ovde vidimo, prvo, odraze svetla koji pripadaju zmaju,

closest to the wall,

koji je najbliži zidu,

followed by bright dots from the disco ball

zatim sjajne tačke od disko kugle

and other reflections from the bookcase.

i drugih odraza od police za knjige.

And finally, we see the reflected waves of light from the statue.

Konačno, vidimo odraze svetlosnih talasa od statue.

These waves of light illuminating the wall

Ovi svetlosni talasi koji osvetljavaju zid

are like fireworks that last for just trillionths of a second.

su poput vatrometa koji traju samo bilioniti deo sekunde.

And even though these objects reflect light differently,

I čak iako ovi objekti drugačije odbijaju svetlost,

we can still reconstruct their shapes.

mi i dalje možemo da rekonstruišemo njihove oblike.

And this is what you can see from around the corner.

Ovo možete videti iza ugla.

Now, I want to show you one more example that's slightly different.

Želim da vam pokažem još jedan primer koji je malo drugačiji.

In this video, you see me dressed in this reflective suit

U ovom videu, vidite mene obučenog u reflektivno odelo

and our camera system is scanning the wall at a rate of four times every second.

i naš sistem skenira zid četiri puta svake sekunde.

The suit is reflective,

Odelo je reflektivno,

so we can actually capture enough photons

tako da zapravo možemo da uhvatimo dovoljan broj fotona

that we can see where I am and what I'm doing,

tako da možemo videti gde sam i šta radim

without the camera actually directly imaging me.

bez da me kamera direktno snima.

By capturing photons that scatter from the wall to my tracksuit,

Beležeći fotone koji se šire od zida ka mom odelu,

back to the wall and back to the camera,

nazad do zida i nazad do kamere,

we can capture this indirect video in real time.

možemo da snimimo ovaj indirektni video u realnom vremenu.

And we think that this type of practical non-line-of-sight imaging

Mislimo da bi ovakva vrsta praktičnog stvaranja slike van vidnog polja

could be useful for applications including for self-driving cars,

mogla da bude korisna za primene, uključujući i samovozeće automobile,

but also for biomedical imaging,

ali takođe i za biomedicinsku obradu slike

where we need to see into the tiny structures of the body.

gde je potrebno da vidimo unutar malih struktura tela.

And perhaps we could also put similar camera systems on the robots

Možda bismo mogli da slične sisteme kamera postavimo na robote

that we send to explore other planets.

koje šaljemo da istražuju druge planete.

Now you may have heard about seeing around corners before,

Možda ste ranije čuli za gledanje iza ugla,

but what I showed you today would have been impossible

ali ono što sam vam danas pokazao bilo bi nemoguće

just two years ago.

pre samo dve godine.

For example, we can now image large, room-sized hidden scenes outdoors

Na primer, sada možemo da obradimo velike, skrivene scene veličine sobe

and at real-time rates,

napolju i u realnom vremenu,

and we've made significant advancements towards making this a practical technology

i napravili smo značajne napretke ka tome da ovo postane praktična tehnologija

that you could actually see on a car someday.

koju biste možda mogli da vidite jednog dana na automobilu.

But of course, there's still challenges remaining.

Naravno, još uvek preostaju izazovi.

For example, can we image hidden scenes at long distances

Na primer, možemo li obraditi skrivene scene na velikim udaljenostima

where we're collecting very, very few photons,

odakle sakupljamo veoma, veoma malo fotona,

with lasers that are low-power and that are eye-safe.

laserima koji su male snage i koji su bezbedni za vid.

Or can we create images from photons

Ili, možemo li napraviti slike od fotona

that have scattered around many more times

koji su se raširili mnogo više puta

than just a single bounce around the corner?

od jednog odskoka iza ugla?

Can we take our prototype system that's, well, currently large and bulky,

Možemo li uzeti naš prototip sistem koji je, pa, trenutno veliki i glomazan,

and miniaturize it into something that could be useful

i minijaturizovati ga u nešto što bi moglo biti korisno

for biomedical imaging

za biomedicinsku obradu slike

or perhaps a sort of improved home-security system,

ili možda na neki način poboljšan sistem kućne bezbednosti,

or can we take this new imaging modality and use it for other applications?

ili možemo li uzeti ovu novu modalnost obrade slike i naći joj druge primene?

I think it's an exciting new technology

Mislim da je to uzbudljiva nova tehnologija

and there could be other things that we haven't thought of yet

i da mogu postojati druge stvari za koje još nismo ni pomislili

to use it for.

da je iskoristimo.

And so, well, a future with self-driving cars

Tako da, budućnost sa samovozećim automobilima

may seem distant to us now --

nam sada može delovati daleko -

we're already developing the technologies

već razvijamo tehnologije

that could make cars safer and more intelligent.

koje bi mogle da automobile učine bezbednijim i inteligentnijim.

And with the rapid pace of scientific discovery and innovation,

A uz ubrzan tempo naučnih otkrića i inovacije,

you never know what new and exciting capabilities

nikada ne znate kakve nove i uzbudljive mogućnosti

could be just around the corner.

se kriju tik iza ugla.

(Applause)

(Aplauz)