AD

Show AD

Deep Learning Computer Data

 Deep Learning Data 



so how do we make computers see like us one way is to use deep learning deep learning is a machine learning technique which allows computers to do what comes naturally to us learning by example deep learning is a key technology behind self-driving cars it enables them to recognize lampposts and to tell the difference between pedestrians and stop signs and it is used in so many other ways in deep learning a computer model is trained using a large set of label data and neural network architecture that contains many layers the term deep refers to a number of layers in a neural network a traditional neural network has about two to three layers but a deep learning neural network has over 150 and can have as many as thousands this is the reason why deep learning neural networks can achieve such a high level of accuracy and can even beat humans at recognizing things I'll go out well deeper into deep learning later on in another Blog but for now let's move on to the five applications on computer vision so the first really cool application is image classification image classification is the process of identifying objects in images although it's really easy for us humans it is an extremely difficult task for computer but with deep learning computers are able to do that some really cool uses of image classification is actually insecurities like identifying faces in CCTV footage and even in real-time at airport another popular example of the use of image classification is in self-driving cars because without image classification self-driving cars would not be possible image classification allows self-driving cars to identify where the road is and what all the obstacles are so it's such a huge part of self-driving cars another example of image classification is in the healthcare industry and how you can you're able to identify tumors in scans of endoscopies and biopsies second application of computer vision that I'm gonna be talking about its image colorization image colorization does exactly what it says it basically adds color to black and white.

 

images however achieving the real color of a black and white image is extremely difficult when we have a grayscale image we essentially only have one kind of information about that image and that is its intensity so how do we determine the color of each of those pixels in that image turns out each of the pixel in a black and white image has 313 different color possibilities to choose from so essentially this is the biggest problem that image colorization is trying to solve so image colorization is used to color black-and-white images especially old black and white images next up is 3d reconstruction for us humans understanding how an object looks like in 3d when we're just looking at it in 2d is pretty easy however this is an extremely difficult task for computers because it's hard for them to look at a 2d image and derive how it would look like in 3d there are several ways in which 3d reconstruction can be I doubt one of the most popular ways is to take several photographs of a singular object at different angles and mesh them together to get a 3d image next up is image synthesis image synthesis is an extremely broad genre essentially image synthesis uses neural networks to generate images which are similar to existing images in our data set but not exactly the same in 2017 by making use of image synthesis researchers managed to generate photographs of human faces which looks super realistic by using photos of celebrities in 2016 another paper release showed how images could be generated from textual descriptions the project use extremely detailed descriptions to generate images of birds and flowers another popular example of image synthesis in the mainstream media is the face app I'm sure many of you have tried it by now the face app uses image synthesis to generate hyper realistic images of us when we are old the last computer vision application that I'm going to be talking about is image style transfer image child transfer is essentially taking two images and blending them together to produce one it is used in so many ways especially in computer graphics and media and many more style transfer is a technique which uses two images the first one is the content image and the second one is the style reference image it takes these two images and blends them together so that the output image looks like the content image but painted in the style of the reference image that's all for today's blog hope you guys found it interesting and hope you guys learn something about computer vision. Thanks for read my blog.

No comments:

Info Upated

Powered by Blogger.