How Computer Vision Applications are Changing the World
Computer Vision is the field of Artificial Intelligence that enables computers to understand and analyze the real world. Using Deep Learning models, machines can now accurately identify and classify objects from within digital images and then react to what they “see.” An increased need for automation and a growing demand for vision-guided robotics and other industry-specific systems are driving massive adoption of Computer Vision applications. The market is expected to grow from USD 10.9 billion in 2019 to USD 17.4 billion by 2024, representing a CAGR of 7.8%.
Origins of Computer Vision
Computer Vision took its first steps in the 1950s, when early neural networks began to detect the edges of objects and to sort them by their shapes. In the 1970s, the first commercial Computer Vision applications were used to interpret written text for the blind, using optical character recognition (OCR). As the internet matured in the 1990s, large sets of images became available online for analysis, driving the development of facial recognition programs. As Computer Vision applications evolved, algorithms were implemented to solve unique challenges across a wide range of sectors, and have steadily become more sophisticated over time.
Today, the ubiquity of smartphone cameras ensures a constant stream of photos and videos and Computer Vision technology has become easily accessible, making it even more appealing to enterprise. Accuracy rates for object identification and classification have jumped from 50% to 99% in less than a decade and today’s systems are even more accurate than humans.
The technology’s impact is being felt across a wide range of fields that rely on computers to analyze images. These include the military, industrial, healthcare, automotive, data and retail domains. As Computer Vision continues to mature, the applications of its technology seem almost endless. Let’s explore a few.
For modern armies, Computer Vision is a vital enabling technology that helps security systems detect enemy troops or saboteurs and enhances the targeting capabilities of guided missile systems. Military concepts such as situational awareness rely heavily on image sensors to deliver battlefield intelligence used for tactical decision-making. Another key Computer Vision application is the areas of autonomous vehicles, which need to navigate challenging terrain and detect adversaries. Computer Vision systems also support human drivers and pilots, enabling them to evade enemy fire. As is the case with many military innovations, technology transfer is now benefiting a broad range of industries.
In manufacturing, the closely allied field of machine vision has long been used for automated inspections, identifying defective products on the production line and for remote inspections of pipelines and equipment. The technology is also used to automate and optimize operational and control processes, by flagging irregular events or inconsistencies. Computer Vision examples in industry include predictive maintenance, product assembly, package inspection, barcode reading for effective tracking, text analysis and control of robotic workers. For example, Osprey Informatics employs Computer Vision applications to monitor remote oil wells in order to eliminate unnecessary technician visits, halving the associated costs.
Since 90% of all medical data is image-based, numerous Computer Vision applications in the healthcare sector have emerged. The technology can detect abnormalities in imagery derived from MRI and CAT scans with a far higher degree of accuracy than medical professionals can achieve. With the ability to detect early-stage tumors, arteriosclerosis and thousands of other conditions, radiologists, cardiologists and oncologists have embraced Computer Vision. It has also become an essential element of many invasive procedures. For example, Gauss Surgical has developed a solution that monitors blood loss in real time by detecting the amount of blood on surgical sponges. The technology is currently being used in operations including caesarian births.
One area which has captured the public’s imagination is driverless cars, which rely heavily on Computer Vision and Deep Learning. While not yet at the point of fully replacing the human driver, autonomous vehicle technology has advanced significantly over the past few years. AI analyzes data gleaned from millions of motorists, learning from driver behavior to automate lane finding, estimate road curvature, detect hazards, and interpret traffic signs and signals. Waymo, for example, has trained its Computer Vision algorithms by driving seven million miles on public roads.
To assist humans with identification tasks and organizing information, Computer Vision tools and Deep Learning models must be trained, requiring huge volumes of labeled data. This is generally carried out by humans, a time-consuming, expensive and often inaccurate process. As Deep Learning algorithms evolve, they are largely replacing the manual tagging process through an approach known as the crowdsourcing of expertise — the automatic real-time collection and tagging of data generated by professionals as they go about their everyday work. Crowdsourced data, a hot topic in fields as diverse as cloud computing and genealogy, is the most efficient, accurate and cost-effective approach to data collection, and is now powering an increasing number of next-generation platforms.
Customer Experience — the Killer App
Retail innovations like Amazon Go have captured the headlines recently, but over the past few years, Computer Vision applications and technologies have been successfully integrated into the CRM domain, from sales and marketing to customer assistance and retention. Computer Vision can be a force multiplier in retail, providing valuable insights into customer behavior and aiding both upselling and cross selling. It can add essential information to a customer’s profile based on visual data from smart telematic devices, a game-changer for insurance and utility companies. It can also help predict issues before they happen, allowing customer care teams to avoid dissatisfaction and churn. When a customer reaches out to a company with a technical or service issue, Computer Vision can effectively route the case to the relevant agent, and help the employee diagnose and resolve the problem much faster than if they were relying on voice or text alone.
Remote Visual Assistance & Self-Service
As consumers progressively embrace smart home devices to suit their lifestyle, entertainment, safety and security preferences, there has been a paradigm shift in the traditional customer assistance model. The IoT is creating new levels of interoperability and complexity, and customers will need more help with installing, operating and maintaining their home device ecosystems. To handle rapidly rising call volumes, companies will therefore have to deliver new levels of self-service, one of the most exciting emerging Computer Vision applications.
Bots with the power of sight can more easily understand the customer’s issue and determine the path to resolution, providing clear and accurate visual instructions in the form of Augmented Reality pointers on the customer’s smartphone screen. The bot can then confirm that the correct actions have been taken and that the issue has been resolved effectively. Relying on massive visual repositories of devices, issues and resolutions, bots powered by Computer Vision and Deep Learning can optimize self-service interactions from start to finish.
Where We’re Heading
The market for Computer Vision applications is growing quickly and as the technology becomes more affordable, we can expect to see a continued rise in the use of Computer Vision image recognition and Deep Learning, helping the dream of smart cities become a reality.
The revolution is well underway. For example, the Department of Homeland Security has recently implemented a biometric monitoring system for the Customs and Border Protection agency to verify travelers at U.S. airports. Meanwhile, the NYC Department of Transportation has begun using image recognition to better understand major traffic events on the city’s busy streets.
This is only the beginning. By recreating humankind’s ability to see, a nearly endless array of Computer Vision applications is rapidly coming into focus.
This article was first published on the TechSee blog.