Tuesday, April 18, 2017

Google Develops Artificial Intelligence to Create Image Sketch

Apparently, Google is currently developing a technology-based capabilities of artificial intelligence (Artificial Intelligence / AI) to identify a sketch. Even the Google AI team is now developing a way so that the computer can draw their own sketches.

Known on 13 April, Google distributed teams 'work' their sketch by relying on a neural network to draw a sketch named Sketch RNN. However, this capability has not been officially released.

Google
AI development team describes the ability to rely on man-made sketches, but does not replicate the overall sketch. Quite the AI capabilities of Google will create their own image.

Because it uses a man-made sketch, a program developed by Google is of course reconstructs the original image using a unique way. They pinned encoder and decoder so that the computer will not remember the actual sketch form.


However, it's not the main goal was to teach the computer to create images of imitation, but rather indicates that neural networks are able to create their own image.



Google AI team, David Ha and Douglas Eck make the process of creation of 70,000 images to teach those skills to the computer system by using a motor sequence. For example, steer and lift the pen.

Although computer-generated picture of a cat is similar to drawing a small child, Sketch RNN able to detect a unique sketch.

"When we sketched the three-eyed cat, computer cat makes a two-edged image. That is, our program has to understand that cats have two eyes," Ha said as reported by Digital Trends.
 

Also Read:

Previous Post
Next Post

post written by: