Technology Trends Digital tools will allow more and more people to have the ability to fake. It won't be long before everyone can use artificial intelligence to perform complex processing on a picture or video.
Smile Vector is a Twitter robot that can generate a motion picture of any star photo smile. If you enter a face picture, it can generate their smiling faces through a deep learning neural network. Although these pictures may not be perfect, they are completely automatically generated. This is another advancement of artificial intelligence in the field of image processing. Maybe soon, the falsification of pictures, sounds and videos will become easy. Imagine if your new version of Photoshop can edit pictures as easily as editing text in Word. Would you still believe in your own eyes?
This will be a technological leap. "I'm pretty sure this will be a technological leap," says Tom White, a developer of Smile Vector, a lecturer at the Victoria University School of Design. "This not only means that we have the ability to modify the image, but also means that it is simple and easy to apply to everyone." White compares his work to the "provocation" of the real world, intended to declare the technical capabilities of artificial intelligence. "I think that people outside of science and technology don't know what machine learning can do. You can imagine that if we add such a filter to Instagram, you just need to choose "laugh" or "smile" and the picture will be processed instantly. Well, everyone can do this on their own mobile phones."
Smile Vector is just the tip of the iceberg of new technology. It is difficult to provide a comprehensive overview of modern artificial intelligence multimedia processing technology, but we can find some interesting applications. For example: create a 3D facial model from a 2D image; use a human "model" to change the facial expressions of the person in the video in real time; change the light source and shadow in the image; automatically generate sound for the silent video; Bald became bald; using video clips to "revive" friends and so on. These examples are only a small part of them.
"This area is rapidly developing," said Jeff Clune, professor of computer science at Wyoming University. "Every month I see new products." Clune's research does not involve modifying images, but generating images directly. He and his team have trained neural networks through object recognition since 2015. The study was based on a neurological study by Quian Quiroga et al. in 2005. They determined that excited neurons are produced in the human brain when faced with certain images, and teaches the entire network to produce images that maximize this stimuli. .
In 2015, they generated images like this:
By 2016, their research has made great progress:
In order to generate these pictures, neural networks need to be trained on a database of such pictures. Once it has learned enough images of ants, red fins and volcanoes, it can generate its own version of the command. The two current bottlenecks are image resolution (these images are no larger than 256×256), and there are enough numbers of marked images to train the neural network. Clune said: "The challenge we face today is not the model itself, but the lack of higher resolution data sets. How long will it take to produce a true full HD image? We don't know, but it should only take a few years. Instead of decades
When these technologies are perfect, they will soon become popular. “Style conversion†is a good example. This app uses neural networks to apply the style of one picture to another. An important paper in this direction was published in September 2015. Subsequently, the researchers of the paper made their research into an open source web application in January 2016. In June, a Russian start-up company pioneered the code improvement and made a mobile app Prisma. This application allows everyone to create different artistic style photos on their mobile phones and then share the created photos with social networks. in. Prisma became the darling of social networking. In November of this year, Facebook released its own version of the style conversion application, adding some new features to Prisma. In less than a year, this technology has completed the process from cutting-edge research to commercial product formation. This is the rhythm of the development of such tools.
Clune believes that artificial intelligence image generation applications will play an important role in the creative industry in the future. Furniture designers can use it as an “intuition pump.†After providing a dataset of chairs for a neural network, they can ask it to automatically generate variants of these chairs. This is an innovative way of creating. Other uses of image generation may generate scenes based on what the user is saying in real time in areas such as video games and virtual reality. Want a dragon? As long as the order is issued, it will be generated. Researchers are already working on these cutting-edge interactions. In the image below, the image on the right is simply based on the left subtitles.
It can obviously be used to engage in mischief. There is a program called Face2Face that transforms the characters in the video into puppets, allowing your expressions to be mapped onto their faces. Researchers used the lens of Trump and Obama to prove this. In a new study published by Adobe (Project VoCo), users can edit human speech, and the company stated that it can be used to adjust pronunciation and dialogue in videos just as Photoshop edits images. You can now create video clips of politicians, celebrities, orators, regardless of content. Then you can publish your clips on any social networking page and see how quickly it spreads across the Internet.
Figure: Face2Face changes the character's mouth shape in real time
This does not mean that machine learning editing tools will make our society no more truthful. After all, the history of human falsification has a long history. The practice of retouching photographs was first carried out in darkrooms. The media often misreport false images as true. From North Korea's "missile launch" to Osama bin Laden's "corpses", these pictures can be seen on the pages of some British tabloids. Similarly, the same is true for video—for example, the 2015 Planned Parenthood scandal, which edited hidden shots to support some sensational misrepresentations.
However, we cannot deny that digital tools will allow more and more people to have the ability to forge. It won't be long before everyone can use artificial intelligence to perform complex processing on a picture or video. Once everyone can process photos as fast as Word, things will become more complicated. Now, it's not difficult to check the forgeries and manipulations made by artificial intelligence. (Blurring is the most common method, and low resolution makes it “look like falseâ€), but researchers have been trying to improve them. technology.
If everyone could quickly and easily modify a photo like a professional, what would happen in this world?
There are more and more fake things in the real world. This is definitely a pleasant thing for the conspiracy theorists, but it will also greatly reduce people’s confidence in the journalism industry. Once people know that forged photographs are circulating in the news industry, even if they see the truth, they will begin to doubt, whatever the reason. (For example, in 2012, Hurricane Sandy's blog picture was confirmed to have forged pictures, but there are also real pictures.) If the new software allows us to handle audio and video content as easily as pictures, this actually weakens it. Another pillar of the media's "true and credible" evidence.
Artificial intelligence researchers in this field have had a direct experience of the upcoming media environment. Clune said: "I'm currently in a real world where I'm dizzy." "People send me some real pictures, but I can't help but think they don't look like fake. When they send me When I forged some pictures, I assumed that these pictures were true because they were really high quality. Gradually, I began to think that we wouldn’t know the difference between true and false. It depends on people’s self. Try and self-learning skills."
Our metal Wire Mesh includes various decorative meshes, welded wire mesh, stainless steel wire mesh, chain link fence, hexagonal wire mesh ect.
Wire Mesh
Decorative Mesh,Decorative Metal Screen Mesh,Decorative Metal Mesh,Metalic Decorative Mesh,Decorative Wire Mesh, Decorative Mesh Screen
HEBEI TUOHUA METAL PRODUCTS CO.,LTD , https://www.penetting.com