Joseph Redmon: How computers learn to recognize objects instantly
Joseph Redmon: Como aprenden os ordenadores a recoñecer obxectos instantaneamente
Joseph Redmon works on the YOLO algorithm, which combines the simple face detection of your phone camera with a cloud-based AI -- in real time. Full bio
Double-click the English transcript below to play the video.
crían que lograr que un ordenador
thought that getting a computer
between a cat and a dog
in the state of artificial intelligence.
en intelixencia artificial.
greater than 99 percent accuracy.
superior ao 99 por cento.
put a label to that image --
thousands of other categories as well.
miles doutras categorías.
at the University of Washington,
na Universidade de Washington,
computer vision models.
modelos de visión artificial.
a prediction of dog or cat,
can ou gato;
specific breed predictions.
of granularity we have now.
que temos agora.
in image classification,
na clasificación de imaxes,
when we run our classifier
o noso clasificador
with a pretty similar prediction.
unha predición bastante similar.
there is a malamute in the image,
we don't actually know that much
en realidade, non sabemos moito
detección de obxectos,
called object detection,
and try to find all of the objects,
de atopar todos os obxectos,
when we run a detector on this image.
un detector a esta imaxe.
algoritmos de visión artificial.
with our computer vision algorithms.
that there's a cat and a dog.
un gato e un can.
algúns datos adicionais.
on top of computer vision,
baseado na visión artificial,
or a robotic system,
ou un sistema robótico,
of information that you want.
que necesitamos.
you can interact with the physical world.
interactuar co mundo físico.
coa detección de obxectos,
on object detection,
to process a single image.
en procesar unha soa imaxe.
speed is so important in this domain,
da velocidade neste ámbito,
to process an image.
en procesar unha imaxe.
it makes predictions,
by another factor of 10,
at five frames per second.
a cinco imaxes por segundo.
like this driving my car.
coma este conducise o meu coche.
running in real time on my laptop.
funcionando en tempo real no meu portátil.
as I move around the frame,
os meus desprazamentos pola imaxe,
of changes in size,
a moi diversos cambios de tamaño,
on top of computer vision.
baseados na visión artificial.
a thousand times faster.
mil veces máis rápido.
object detection systems
de detección de obxectos
on each of these regions,
en cada unha desas rexións,
detections in the image.
thousands of times over an image,
miles de veces sobre unha imaxe.
to produce detection.
para producir unha detección.
to do all of detection for us.
para que fixese todo o proceso.
and class probabilities simultaneously.
e as probabilidades de clase.
at an image thousands of times
unha imaxe miles de veces
the YOLO method of object detection.
de detección de obxectos.
we're not just limited to images;
limitados ás imaxes estáticas;
that cat and dog,
ver o gato e o can,
and interact with each other.
like spoon and fork, bowl,
como culler e garfo, cunca,
out into the audience
our threshold for detection a little bit,
o limiar de detección,
out in the audience.
eses sinais de stop.
is happening in real time
estase realizando en tempo real
object detection system,
de uso xeral,
para calquera ámbito.
células cancerosas
already using this technology
que xa están a usar esta tecnoloxía
like medicine, robotics.
como a medicina ou a robótica.
of animals in Nairobi National Park
do Parque Nacional de Nairobi
of this detection system.
do sistema de detección.
free for anyone to use.
poida usalo libremente.
even more accessible and usable,
máis accesible e práctica,
of model optimization,
a optimización de modelos,
running on a phone.
funcionase nun teléfono.
now we have a pretty powerful solution
unha solución bastante potente
de visión artificial de baixo nivel,
and build something with it.
para facer algo con ela.
with access to this software,
con acceso a este software,
will build with this technology.
construirán con esta tecnoloxía.
ABOUT THE SPEAKER
Joseph Redmon - Computer scientistJoseph Redmon works on the YOLO algorithm, which combines the simple face detection of your phone camera with a cloud-based AI -- in real time.
Why you should listen
Computer scientist Joseph Redmon is working on the YOLO (You Only Look Once) algorithm, which has a simple goal: to deliver image recognition and object detection at a speed that would seem science-fictional only a few years ago. The algorithm looks like the simple face detection of a camera app but with the level complexity of systems like Google's Deep Mind Cloud Vision, using Convolutional Deep Neural Networks to crunch object detection in realtime. It's the kind of technology that will be embedded on all smartphones in the next few years.
Redmon is also internet-famous for his resume.
Joseph Redmon | Speaker | TED.com