Google AI can recognize scribbled graffiti

The Verge Chinese Station reported on November 17

The artificial intelligence lab created by Google recently released a new AI Web service called "Quick, Draw!"

This is a drawing recognition service that can accurately identify scribbled graffiti and even multi-lingual languages ​​including Chinese in just 20 seconds, further demonstrating the future of AI in the areas of self-learning, key image and character recognition.

"Quick,, Draw!" differs from traditional mapping software in that when you start painting, the AI ​​starts to recognize and automatically emits a voice to describe the object you are painting. When the drawing is nearing completion, its sound becomes more positive and the object is correctly identified, which is very close to the process of human guessing.

The impressive part is that the technology's ability to recognize objects is very broad. For example, if you draw a cat, 100 people may have 100 paintings. AI is not just a traditional database to add recognition, but it can be used in the rapid painting process. In the context of association and judgment.

If you want to draw a speed boat but it looks like a half moon, but there are wavy lines that represent water below, AI can use wavy lines to judge that it is a boat; at the same time, because the speed boat has no sail, AI can further affirm It is a speed boat, not a sailing boat. In addition, if it goes wrong, the neural network mechanism will let it learn in error and reduce the error rate.

If you are interested in this project, you may wish to go to its official website to see, there are also some other AI experiments Google investment, are very interesting. (Original author: Nick Statt compilation: Ben)

Click to read the original English

【The relevant rights in the Chinese language of The Verge's works in the United States are owned by Tencent Corporation. Without authorization, they may not be reproduced or excerpted. 】

Solar Panel

Solar Panel,Poly Solar Panel,Mono Solar Panel,Solar Panel For Power Stations

Fuzhou Mei Li Cheng Imp&Exp Co., Ltd , https://www.mlc-solar.com