The algorithm that converts a two -dimensional image into a "high -precision 3D image" accelerates the evolution of AI
delivery
2コメント2件Illustration: Elena Lacey; Getty Images
The beginning of the big swell surrounding the current artificial intelligence (AI) technology dates back to 2012. This year, an academic contest was held to compete how accurately the algorithms could recognize the objects in the photos. AI, which can answer ethical questions, and their ability and limit researchers, built an algorithm, with a rough hint in a mechanism for human neurons to respond to new information. He discovered that thousands of images were loaded into the algorithm, which could dramatically enhance recognition accuracy. This groundbreaking discovery has been causing severe earthquake in the academic research and business world, causing changes to many companies and industry. And now, by training the same kind of AI algorithm, a new technology that turns the two -dimensional (2D) image into an expressive 3D image has appeared. This new technology, which makes computer graphics (CG) and the two worlds of AI, have a great power to completely change the way of imaging games, virtual reality (VR), robot engineering, and cars. Some experts think that this technology may become smarter and acquire more intelligence as humans, understand everything in the world, or discuss the reason. is.
A revolutionary technology with many uses
"It is a technology that is now attracting much attention in the hot topic," says Ken Goldberg, who studies robot engineering at the California University Berkeley School (UCB). He is using this technology to enhance the abilities of robots operating with AI and make it an unfamiliar shape. He says that this technology should have hundreds of uses from entertainment to architecture. What is used in this new technology is a technique called "neural rendering" that reads several 2D snapshots using a neural network and generates a 3D image. The interest in this technology, which was born from the fusion of various concepts over CG and AI, has suddenly increased after 20 years. In April of this year, the UCB and Google joint research team showed that only a few 2D photos on the neural network could reproduce the same scene in a very realistic 3D image. This algorithm, which uses the movement of light passing through the atmosphere, is designed to calculate the density and color of each data point in 3D space. This makes it possible to convert the 2D image to a real 3D image from anywhere. The core neural network of this technology is the same as an image recognition algorithm that analyzes the pixels of 2D images published in the 12 -year academic contest. The newly created algorithm converts 2D pixels into a three -dimensional pixel called "Voxel". The video introducing the technology named Neural Radiance Fields (abbreviated Nerf) has made many researchers. "I've been involved in computer vision research for 20 years, but when I watched this video, I said," No, this is amazing, "said Georgia Institute of Technology's professor Frank Delatt. Everyone who researches computer graphics must be recognized for the innovation of this technology. To complete a real and detailed 3D image, you usually need to be troublesome for hours. However, with the emergence of the new method, it became possible to create such a 3D image from just a few minutes. It is also possible to synthesize images in an unprecedented way or move the finished synthetic image freely. "It's a very original and valuable technology. It's extremely unusual to give such a compliment to the technology of just two years after birth," says Delat. Delating says that there are many ideas born after this new technology. For example, a move selfie that looks like a person's head around a few still images (also called "nerfie" after the name of this new technology), creates a 3D avatar from one photo, There are various ideas, such as how to automatically change the lighting of the scene. And Nerf has penetrated the industry at an amazing speed. One of the Nerf researchers, now Google, Ben Mildenhole, describes the rise of the current research and development as "a slow, large wave."
次ページは：メタヴァースの景観が変わるPage 1/2
最終更新:WIRED.jp
