Joseph Redmon: How computers learn to recognize objects instantly
Joseph Redmon: Cara komputer belajar mengenali objek secara langsung
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.
thought that getting a computer
merasa bahwa memerintahkan
between a cat and a dog
in the state of artificial intelligence.
dalam bidang kecerdasan buatan.
greater than 99 percent accuracy.
dengan akurasi lebih dari 99 persen.
put a label to that image --
beri label gambar itu,
thousands of other categories as well.
ribuan kategori lainnya.
at the University of Washington,
di University of Washington,
computer vision models.
model penglihatan komputer.
a prediction of dog or cat,
prediksi anjing dan kucing,
specific breed predictions.
secara spesifik.
of granularity we have now.
prediksi lebih detail lagi.
dari jenis malamute.
in image classification,
dalam klasifikasi gambar,
when we run our classifier
with a pretty similar prediction.
prediksi yang lumayan mirip.
there is a malamute in the image,
dalam gambar,
we don't actually know that much
kita tidak tahu banyak tentang
called object detection,
disebut deteksi objek,
and try to find all of the objects,
mencoba mencari semua objek,
when we run a detector on this image.
dijalankan pada gambar.
with our computer vision algorithms.
penglihatan komputer.
that there's a cat and a dog.
tahu ada kucing dan anjing,
on top of computer vision,
berkekuatan penglihatan komputer,
or a robotic system,
atau sistem robotika,
of information that you want.
yang Anda inginkan.
you can interact with the physical world.
dapat berinteraksi dengan dunia fisik.
on object detection,
to process a single image.
memproses satu gambar.
speed is so important in this domain,
sangat penting dalam domain ini,
to process an image.
untuk memproses 1 gambar.
it makes predictions,
komputer memprediksi,
by another factor of 10,
hingga 10 kali lipat,
at five frames per second.
lima bingkai per detik.
like this driving my car.
mengemudikan mobil saya.
running in real time on my laptop.
laptop dalam waktu nyata.
as I move around the frame,
saya bergerak di sekitar bingkai,
of changes in size,
berbagai perubahan ukuran,
on top of computer vision.
penglihatan komputer.
a thousand times faster.
seribu kali lebih cepat.
object detection systems
sekelompok area
on each of these regions,
pada masing-masing area,
detections in the image.
thousands of times over an image,
deteksi pada satu gambar,
to produce detection.
untuk menghasilkan deteksi.
to do all of detection for us.
untuk melakukan semua deteksi.
and class probabilities simultaneously.
sekaligus probabilitas kelas.
at an image thousands of times
satu gambar ribuan kali
the YOLO method of object detection.
metode deteksi objek YOLO.
we're not just limited to images;
kita dapat memproses tidak hanya gambar,
that cat and dog,
melihat kucing dan anjing,
and interact with each other.
bergerak dan berinteraksi.
like spoon and fork, bowl,
sendok dan garpu, mangkuk
out into the audience
our threshold for detection a little bit,
ambang pendeteksinya sedikit,
out in the audience.
lebih banyak penonton.
is happening in real time
object detection system,
sistem deteksi objek,
pada domain gambar mana pun.
atau pejalan kaki,
menemukan sel kanker
already using this technology
yang sudah menggunakan teknologi ini
like medicine, robotics.
obat-obatan, robotika.
of animals in Nairobi National Park
di Taman Nasional Nairobi
of this detection system.
bagian dari sistem deteksi ini.
free for anyone to use.
gratis untuk siapa saja.
even more accessible and usable,
lebih mudah diperoleh dan berguna,
of model optimization,
pengoptimalan model,
running on a phone.
berjalan dalam ponsel.
now we have a pretty powerful solution
sekarang ada solusi yang cukup kuat
komputer level rendah,
and build something with it.
membuat sesuatu dengan memakainya.
dan orang-orang
with access to this software,
perangkat lunak ini,
will build with this technology.
mereka buat dengan teknologi ini.
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