This morning, another AI partial door application broke out! Use algorithms to "take off" women's clothes directly!
Picture from Baotu Network
It seems that the previous face-changing technology can no longer meet the needs of some people, and some people have come up with even more shocking applications. It can no longer be called a side ball application.
DeepNude
This application, called DeepNude, is said to only need to give it a picture of a woman, and the app can automatically "take off" the clothes on the woman with the help of neural network technology, exposing the nude. For users, no technical knowledge is required, and nude photos can be obtained with one click.
At present, DeepNude's official website has been launched, not only has a paid version of $50, but also a free version.
Related downloadable Windows and Linux applications are being launched. The free version will have a watermark on the picture, and the paid version will have a "Fake" mark in the upper left corner of the picture.
And how is the processing effect? The American media Motherboard tested dozens of photos and found that if you input the photos of the "Sports Illustrated Swimsuit Special", the nude photos obtained are the most realistic.
However, this technology has disadvantages. After most photos (especially low-resolution photos) are processed by DeepNude, the resulting image will have artificial traces; while inputting a cartoon character photo, the resulting image is completely distorted.
In other words, through this application, you can get any "nude photos" of any woman. Even though this is just forged by the neural network, it looks very real to the naked eye.
And in this case, the so-called "FAKE" watermark on the Internet is also easy to remove. Once this technology is widely used, it will cause greater harm than the previous pornographic face-changing technology.
Technical Principle
DeepNude is equivalent to the evolved version of DeepFake before face change. Mainly based on an open source algorithm called pix2pix proposed by the University of California, Berkeley in 2017.
Pix2pix uses a Generative Adversarial Network (GAN), which uses a large number of image data sets to generate more realistic images. This algorithm is used in scenarios such as autonomous driving.
Specifically, Pix2pix learns the characteristics of some images through neural networks, and then expands them to other images or videos.
DeepNude used more than 10,000 nude female photos for training and finally achieved the effect of "undressing". In fact, on the new picture, a female nude image is generated, which is false.
The creator of the software said: "The neural network is multiple, because it needs to complete multiple different tasks for a picture.
For example, find the clothes in the image and cover the clothes. Infer the specific body parts and render the image... He said that because of the more tasks, the processing speed will be slower, but he hopes to improve in the future. "
DeepNude can only make photos of women at present, because such photos are easier to collect, and he also wants to be a male.
Widespread controversy
Once this technology was reported on the Internet, it immediately sparked heated discussion. The first reaction of most netizens is that the application of this technology is too vulgar!
There are also many people who have doubts about technology. Is AI technology used to do these things?
It seems that since DeepFakes was made, the Pandora's Box of this AI world has been opened. Each of us can become a victim of pornographic revenge.
In addition, technology discriminates against women. From changing faces in small videos to undressing now, these vulgar applications are all aimed at women. Why don't men become the protagonists of nude photos and small videos?
We have always been talking about technology for good, but it turns out that in the face of human nature, we cannot control the direction of such a weapon as AI. Now that we change our faces and undress, what more terrifying things will happen in the future?
It is hoped that the relevant departments can introduce relevant policies and do not let technology become a sharp edge to stab ourselves.