Is it so floating now?

Editor’s note: This article is from WeChat public account “Brain pole body” (ID: unity007), author Xiaohao. Authorized to reprint.

Some people with common sense know that in the case of detoxification, physiological dependence is easy to get rid of, but heart addiction is difficult to remove.

General detoxification agencies often refer only to the period of time for drug addicts to avoid drug exposure. Three or six months, or one or two years, during which the drug addicts are considered to be successful in detoxification if they perform normally. Because the heart addiction has not been removed, many people will soon be re-addicted with drug addiction and have to re-detoxify, and even repeated.

There is a diagnosis that relies only on physiological diagnosis and lacks a set of psychological drug addiction, which is obviously an important reason for causing drug addiction for a long time.

And now, a diagnostic method that can determine whether a drug addict is successful in psychological detoxification has emerged.

The distance becomes technology

According to the Hong Kong “Sing Tao Daily” report, the Hong Kong University of Education recently applied “eye tracking” technology to drug rehabilitation research for the first time. By judging and analyzing the time and location of drug addicts’ attention to drugs, it can be determined whether they have completely quit “heart addiction”, and the correct rate is as high as 90% or more.

Don’t say what the technical ability behind this news is. The word “eye tracking” alone is enough to make people “eye-catching” and return it to everyone’s vision. So, what kind of situation is “eye tracking” in today? What is it mainly used in? What about the prospects?

For these questions, let’s try to make some answers.

VR: The best partner for eye tracking

Eye tracking technology, as the name implies, captures the information of eye movements through image acquisition equipment, and realizes tracking and analysis of eyeballs, thus serving specific scenes, such as psychological analysis, electronic device screen switching, and even The device is controlled. It is not difficult to see that eye tracking includes three steps: one, the camera, infrared and other acquisition equipment captures the eye movement; second, the terminal comprehensively analyzes and judges the information such as the eye movement trajectory; Decision or suggestion of willingness.

This makes it easy to think of using this technology to implement a screen or e-book flip.Page, and this is what I most wanted when I was young. In fact, it is now a relatively low-level application of eye tracking technology. As early as 2017, Microsoft also built this technology into Windows 10.

The distance becomes technology < /p>

But Windows 10 seems to be a bit more sloppy. It was really sci-fi when watching demo typing, but the practicality was too bad, far less than the typing efficiency of the manual, coupled with the lack of other applications, the eye tracking finally ended up for Microsoft to buy products. The gimmick.

But from this case that can already be seen as a failure, we can have a slight peek: eye tracking is good, no application. And all kinds of technology that can make the eye tracking technology shine, maybe VR. Tobii also regards VR as the best carrier for its eye tracking technology, and launched Pro Glasses 2, a device that adapts to VR helmets. However, a pen-sized thing is used in equipment such as VR helmets. space.

The distance becomes technology

Another problem is that current VR products often render all the scenes when presenting images of games or other applications. The places where the eyes see are clear and unseen, and the places that are not visible are unambiguous.

We can clearly see three drawbacks: it is easy to cause dizziness; it is a waste of calculation for scene rendering other than eye observation, and it also burdens device calculation; the computing power increases, the size of the device is naturally smaller than Where to go. Do you want to make the hardware bigger? That helmet is even more inconvenient.

And eye tracking, can be used to solve this problem.

Tobii developed a retinal center rendering technology, which uses the tracking of the eyeball to render the area of ​​the human eye in high definition, and blurs the area outside the eyeball, which is in line with the general view of the object. Features.

Tobii’s ideas are very correct, but the actual results are limited. After all, it’s image processing, Tobii is obviously not very good – wait, when it comes to image processing, is there a more suitable method than AI?

AI: The clear direction of technology travel

SinceSince the acquisition of Oculus in 2014, Facebook’s efforts in VR have not been reduced, and the money is not worth the cost. These inputs have also been called “paying for the future”. The stone thrown into the sea by Facebook has not sunk, and it has also stirred up a lot of waves. For example, VR image rendering based on eye tracking technology.

In May of last year, Facebook launched the Half Dome prototype, which integrates an eye tracking camera, a wide field of view optical system and an independent zoom display, in order to achieve the same effect as the human eye: from the line of sight The area is blurred to achieve an immersive experience.

Similarly, this prototype was released at the same level as Tobii. Due to hardware and algorithm limitations, the realistic level of rendering is hard to judge as a good experience. In December last year, Facebook matched the artificial intelligence rendering system Deep Focus.

So, what is the difference between the image rendering of Deep Focus and the Tobii and Half Dome prototypes? One word to answer is: natural blur.

Blurring the picture outside the gaze point is the direction, but there are many questions: Is the transition between the clear area and the blurred area smooth and natural? Is the transformation of the image from blurry to clear during the movement of the fixation point natural?

The unnatural transition is like this:

The distance becomes technology

The natural transition is like this:

The distance becomes technology

Obviously the image below has a more natural gaze edge blurring than the image above. This has also become the direction of Deep Focus efforts. And if you simply use the power to solve the problem, then the volume itself exists.The limited VR head display has again made a request. So the researchers of facebook reality lab chose another direction: using deep learning to achieve natural blur. To this end, the research team needs a lot of data to train the artificial intelligence system. But the question is where to find so much data?

They randomly generated scenes composed of a large number of objects, and then used Deep Focus to learn the textures and other features of these objects to generate realistic rendering blur effects. After nearly a year of training, Deep Focus has been able to make a pretty good rendering performance.

Using the latest artificial intelligence technology to track research and make real-time image processing seems like a very good idea. For Facebook, perhaps it is just a measure to enhance the VR immersive experience; and for the entire application of eye tracking technology, the greater significance of artificial intelligence is to provide an effective research direction.

Eruption: Breaking through obstacles and applying the future of the scene

We will return to the eye tracking technology itself. Entering 2019, with the 5G network commercialization as a symbol, more and more voices believe that the large bandwidth and low latency characteristics of 5G will boost the revival of VR and even the outbreak of the industry. Whether it can be modified with “bursts” seems to have become one of the indicators to measure the value of a technology or industry application.

So, for eye tracking technology, whether it is the traditional hardware manufacturer Tobii or the artificial intelligence company represented by facebook, it seems that it has not stopped its technological evolution for a moment. We also apply: When the AI ​​enters the market, is the application of eye tracking technology also exploding?

The distance becomes technology

The following may be the issues that must be considered on the eve of the so-called “outbreak”.

· Terminal computing performance that supports immediate feedback. Tracking, information capture, analysis, and judgment of eye movements need to be completed in an extreme time. Facebook’s Deep Focus uses AI to achieve a natural blur of high-level applications, but it is built on four high-performance GPU computers. On the basis of this, if you want to successfully carry it on a VR helmet or even an electronic device such as a mobile phone or a watch, you must balance the volume and calculation performance. Or, develop a more powerful algorithm.

·There is an exact exclusion of “invalid” eye movements. People’s eye movements are not all intentional, howExcluding unconscious eye movement disturbances, the next step is to focus on research.

·The large-scale expansion of the application ecology. Today, it seems that in order to obtain a large-scale promotion and application of a new technology, in addition to hardware-level matching, there must be an order of magnitude of application support, especially for applications that are just needed or manufactured. VR is very representative, but only VR is obviously not enough, let alone VR itself is in the process of technological evolution. Then, open up more application possibilities for eye tracking technology, and should also be synchronized with the technical research on it.

For example, combining eye tracking with brain-computer interface, through the capture of invisible brain waves and the tracking of external eyeballs, to determine the intentional tendency of the brain; to provide medical help for people with speech impairments, mobility problems, such as typing , call for help, etc.; or used in the car driver fatigue monitoring, car machine interaction, become a third way of interaction outside the finger, voice, increase the intimacy of human-computer interaction and so on.

The distance becomes technology

It can be seen that under the premise of the maturity of the solution, eye tracking actually has the potential to become a “explosive model” of technology: it has a vast application scene. However, hard demand has not yet formed, and the support of technology is still weak. Eye tracking technology may still be in a tepid state for some time to come.

But the development trend of technology is always beyond our imagination. The AI ​​concept is only 60 years after it ushers in a real explosion. Perhaps the god who rules the evolution of human technology will accidentally take a nap, and eye tracking will jump out. Slowly swallowing the track, from the shortcut to the moment to embrace the world.