Using 5G to Bring Revolutionary Breakthrough in Image Recognition

With the roll-out of new technologies comes an abundance of excitement and hype. There is a hope for a better world where life is made more accessible by these technologies. 5G is one such anticipated thing. The introduction of 5G for business is forthcoming. It is an exciting time for companies worldwide who have heard about the many possibilities it can offer.

Contrary to popular belief that 5G will arrive all at once, it shall come in phases. Ericsson's Mobility Report predicts that 5G coverage is expected to reach between 55% and 65% by the end of 2025, on a global scale. The latency target built-in for 5G is 1 ms. And in comparison, video streaming currently experiences a 1,000 ms latency. Far higher!

Due to the high-speed network of 5G, it can boost Artificial Intelligence to new altitudes. As AI and 5G complement each other, businesses expect to see new possibilities that could not be imagined before. This means one can expect considerable investments to bring a wave of multi-billion dollar infrastructure expenditure. Hence the telecom operators may need to level up rapidly to make the most of the billions spent on 5G wireless spectrum licenses.

As 5G proliferates, so will its applications. When integrated with distributed cloud in the network, deploying applications can be more local and closer to end-user. 5G can also enable contextual awareness for Voice-activated assistants, making them more powerful. Along with edge computing, 5G can open up avenues for more extensive information flow on a real-time basis. But the most exciting aspect is image recognition.

In 2017, Intel and Foxconn demonstrated how facial-recognition feature could help to make payments. Intel's Multi-access Edge Computing (MEC) would use this pay via face identification to complete the payment authentication process in 0.03 seconds. This could mean a lesser risk of personal information leakage and minimal credit card fraud.

We have had been using 2D facial recognition systems for over three decades. Although due to technical upgrading, these systems achieved low error rates in controlled environments, yet are quite sensitive to light exposure, pose variation, make-up, and facial expressions. Therefore this led to the arrival of 3D imaging which is far more accurate than the previous ones. Although such cameras use Wide Dynamic Range (WDR), the surveillance centers need to process large volumes of them at a back-end edge data center at a faster speed to provide real-time insights. Therefore 5G will be an ideal solution to this nightmare.

Due to high-speed connectivity and low latency, the distribution of image feeds to the local edge data center can cut the burden on camera network. This is because only results from the image analysis get transmitted via the network. And this too can happen when an operator center receives system alerts. Other than saving in network bandwidth, this also means that the time needed for the analysis is short.

This intriguing feature has plenty of practical functions—for example, traffic safety and surveillance. Cameras positioned at strategic locations can identify instances of illegal parking, using horns are prohibited areas (red light crossing, railway track, etc.), pedestrians and commuters not obeying traffic rules and misconduct. It can also monitor traffic conditions, missing license plate of vehicles, check if bike riders are wearing helmets, locate hazard zones on roads and flyovers, and much more.

Also, blessed by 5G, we can have better video streaming quality too. Infineon Technologies recently made a 3D ToF sensor technology that uses the REAL3 3D Time-of-Flight (ToF) sensor—thus enabling video bokeh function for the first time in a 5G-capable smartphone for optimal image effects. They achieved this feat in collaboration with the patented SBI (Suppression of Background Illumination) technology from PMD which offers a wide dynamic measuring range for any lighting situation, from bright sunlight to dimly lit rooms. It can thereby reduce the loss of data processing quality.

At retail and shopping outlets, it can provide better engagement with customers than clueless and pestering assistants. Further, it shall prevent logjams at checking counter, or checkout-free shopping. At boutiques or apparel stores, it can provide a personalized experience for users by studying their previous shopping behavior data and serve them with an image of how a specific item of clothing would look on them. Similarly, it can track how passersby interact or respond to advertisements in the form of standees, billboards, and so on. Using these demographics based data, advertisement companies can come up with better planning and production value marketing ideas at different locations and times.