The Tianmouc vision chip is capable of processing images at record speeds, up to 10,000 frames per second.
A team of scientists from Tsinghua University in Beijing, China, said they have developed the world’s first brain-inspired vision chip, giving machines human-like visual perception capabilities. People, Interesting Engineering reported on May 31. New research published in the journal Nature.
AI visual perception is laying the foundation for a groundbreaking technological revolution, especially in automated systems such as self-driving cars. However, achieving efficient, accurate, and flexible visual perception in diverse, fluctuating, and unpredictable environments is a major challenge.
The new chip, called Tianmouc, is introduced as the world’s fastest vision chip, with record image processing speed. It collects image information at up to 10,000 frames per second, with 10-bit accuracy, 130 decibels of dynamic range. In addition, Tianmouc also reduces bandwidth by 90% and maintains low power consumption.
Tianmouc is inspired by the human visual system. It analyzes visual information along two pathways: one for perception and one for quick response.
“This is a perception chip, not a compute chip, based on our original technical roadmap. First, it balances speed and dynamic performance in vision chips, and introduces a method “The new computation is different from existing machine vision strategies. Second, this approach mimics the dual pathways of the human visual system, allowing for efficient decision making,” said project team leader Shi. Luping, a professor at the Center for Brain-Inspired Computing Research (CBICR) at Tsinghua University, said.
The research team affirms that Tianmouc will open new doors for advances in the fields of autonomous driving and defense, and can lead to the development of many new applications.