The magic of NVIDIA, a few photos of the 3D Polygon
In order to make 3D polygons with only 2D photographs, designers had to invest dozens of hours. Based on a plane photo, we had to make a look at how to see it at a specific angle, and I had to call the texture, and according to the designer’s skills, my real picture and distant result came out.
As the AI age comes, this series of processes have been able to be in charge of AI. After the artificial intelligence is analyzed in the same space, the artificial intelligence is analyzed and completes 3D modeling based on the actual photo. However, it is a work of hardships that require creativity and imagination, and the AI had to take a considerable amount of time to analyze and reproduce it.
Nerf technology, which is introduced through NVIDIA (Local Time) and NVIDIA GTC session, has improved the 3D modeling rate of these AI or more than 1,000 times. This technique ends 3D modeling in just a few seconds if there is only a dozens of 2D pictures. High-speed neural network training technology was applied to improve speed. The rendering technology that estimates the depth and shape of things with only partials is to take several tens of minutes from the previous time, but applying NERF technology in the NVIDIA Tensor Cores environment to complete it.
In fact, according to the video of NVIDIA, Andy Warhol has created a real-life character that exists in the actual 3D space as a few photos of the parody model holding the Polaroid Camera. In the video, it is depicted as a photo implemented as a four-sheet photo, but in practice, there is a need for a dozens of images taken at various locations, and 3D modeling can be blurred when the movement difference between each picture is severe.
According to the NVIDIA side description, this technology also trained not only the virtual world implementation such as architecture and entertainment, metaverse, but also to train the size and shape of the actual object with the photographs and images of the actual object It can also be used.