Deep learning is exciting because it
learns these complex features of images. And as we discussed earlier. They've had tremendous impact over
the recent years in a variety of computer vision applications. Let me show you a couple
of early examples. So, on the top of the slide here,
what you see is an example of identifying traffic signs based on neural networks. So these are a data of
German traffic signs and the idea is for every image,
identify what sign it is. And they were able to get 99.5%
accuracy using a deep neural network, which is pretty cool. On the bottom there, you see an example
that came out of some work from Google on identifying the house numbers based
on what's called Street View data. This is the data that Google
uses driving around cars and photographing all sorts of
streets around the world. And you see the images are pretty complex,
and still they're able to get 97.8%
accuracy on the per character level. Now these were exciting results. But the one that changed everything. The really excited field happened in 2012. So for many years, there was a image
competition called ImageNet. And in 2012, the ImageNet
competition included 1.2 million training images from about
1,000 different categories. And the idea was can you
classify this image. Is it not just a dog, but
is it a golden retriever or a labrador? Very, very fine level detail. Now there's many teams competing. These are the top 3 teams. A team called OXFORD_VGG,
which got pretty decent accuracy. So if you look at their top five
guesses can you get the right thing in five guesses. They were getting only about 25% error. There's a thing called ISI
that did a little bit better. And those are using traditional
techniques like the SIFT[1], a little bit more elaborate,
kinda like that. Now that year,
is a team called SuperVision. That team used a deep, neural network and
had a huge gain over the competitors and that performance really sparked a lot of
excitement of using deep neural networks in computer vision because if they
have to just use hand coded features, you'd learn them automatically. Now that neural network that won the
competition with the SuperVision team was called the AlexNet neural network and I'm
showing here an image from their paper, and that neural network involved 8 layers,
60 million parameters, and was only possible because of new
training algorithms the could deal with lots of images and
lots parameters, and the GPU implementation that would
really scale to large data sets. [MUSIC]