Why is it said that machine vision is the next frontier of artificial intelligence?

Abstract: Venturebeat, a foreign science and technology website, published an article stating that artificial intelligence has been weakly developed in the previous year, bringing more and more advantages to people. In the future, machine vision will be the next frontier category of artificial intelligence. Following the development of this type of technology, new artificial intelligence-driven devices will be presented next year.

Why do machines need vision? Vision is the primary sense organ. For machines to understand humans and provide them with the support they need, they must be able to investigate and perform in the field of vision. The detailed method can be a small camera that helps blind people "see" and understand the world around them, and perhaps a home monitoring system that can accurately distinguish drifting cats, moving branches, and thieves.

Reasonable electronic equipment is becoming more and more important in people’s daily life. We have also found that more and more equipment applications fail because they do not have satisfactory and robust visual functions. For example, if the drone crashes in the air, the robot vacuum cleaner cannot absorb it. Something to suck.

Machine vision is a rapidly expanding branch of artificial intelligence, which aims to give machines a vision comparable to humans. As researchers use specialized neural networks to help machines recognize and understand images of the practical world, machine vision has made great strides in the past few years. Today's computers can do a variety of tasks in visual recognition, from identifying cats on the Internet to identifying specific faces in many photos. However, this type of technology still has a long way to go.

At that time, machine vision was out of the data core and used for various purposes, from autonomous driving of drones to food cleaning.

The basic image classification is now much simpler, but in extracting the essentials and information from the messy scene, the machine is facing a series of new problems. The illusion problem is a good example of the long road ahead for industrial vision inspection machines.

For example, when people see two generalized images of faces facing each other, they see more than just abstract shapes. Their brains will perform further interpretations, allowing them to recognize multiple parts of the image, see two faces, and perhaps a vase.

But for machines, such images are very difficult to understand. The basic classifier cannot distinguish between the two faces and the vase. It will see objects such as hatchets, hooks, bullet-proof clothing and even guitars. The system cannot recognize that those objects are beside the image, which shows that the recognition of such images is challenging for machines.

In addition, just like messy images, the world of practice is also very messy. Flying normally by the side is not something that can be done just by developing an algorithm to analyze the data. It requires a clear understanding of the actual scene, and then be able to act accordingly.

Robots and drones face many such obstacles, and defeating these challenges is a top priority for those participating in artificial intelligence innovation.

As technologies such as neural networks and specialized machine vision hardware continue to expand, the gap between machine vision and human vision is rapidly shrinking. Soon, there may even be robots with better vision than humans. They can perform all kinds of messy missions and can operate completely automatically. (Haohui)

Venturebeat, a foreign science and technology website, published an article stating that artificial intelligence has been weakly developed in the previous year, bringing more and more advantages to people. In the future, machine vision will be the next frontier category of artificial intelligence. Following the development of this type of technology, new artificial intelligence-driven devices will be presented next year.