WORD day! Yahoo Open Source Network

WORD day! Yahoo Open Source Network
GuideIf you are interested in how NSFW (No Safe For Work, don't watch it during Work hours) tags on the Internet are classified, you should be interested in the news, yahoo has just opened up a deep learning neural network built on its own to automatically detect pornographic content in images.

Although it is simple, it is indeed difficult to use machines to distinguish pornographic images. After all, it is difficult to distinguish between pornographic images in legal terms. It can even be said that obscene content cannot be defined at all. The former Justice of the United States, port Stewart, has a famous saying: I saw it, only to know whether it is. A machine has no passion. Any image is just a file in its view.

After years of research and practice, convolutional neural networks have proved to be the best image classification tool. As long as a database is provided, algorithms can learn the characteristics of such images, such as photos of classification dogs, you only need to recognize the tail and nose head, classify the car, and give the front-face door handle to the machine "desire". Make up your own brain.

Yahoo's porn image recognition system will give the image a NSFW value, the most harmless is 0, the most extreme is 1, The NSFW Value of the girl spent in the question map is 0.001, iron three sisters wear a little exercise, and the value increases to 0.116, which is not the standard of pornographic images.

After the open source, this recognition system can be used in many occasions. In addition to the review system that can be imagined in the first time, for example, it can also be used as an automatic NSFW tag, it helps you filter out outdated images in your work mailbox (from other employees ).

If you are interested, you can go to Yahoo's article (https://yahooeng.tumblr.com/post/151148689421/open-sourcing-a-deep-learning-solution-for) to view details, the model can be downloaded on GitHub (https://github.com/yahoo/open_nsfw/blob/master/README.md), if you want to build your own porn image recognition system, you must provide your own data, but I think this is not a problem.

(Image placement comes from the network, not the original image)

From: https://linux.cn/article-7827-1.html

Address: http://www.linuxprobe.com/yahoo-detection-picture.html #