From Damage Assessment to Driver Assistance, Computer Vision Tech Pervades
In a historic 2012 ImageNet competition, a team led by Geoff Hinton debuted a new network architecture whose performance outshone all previous efforts at computer-based image recognition. With that, the era of deep learning was born, with computer vision as its original use case. Since then, computer vision capabilities have raced ahead.
In layman terms, computer vision (CV) is the automation of human sight. The ability to automate sight opens up market opportunities across the economy, including insurance which relies on visual assessment of assets: to price and underwrite policies or to determine degree of damage after an accident. Computers are primed to see via exposure to newer objects and closed loop feedback on previously identified objects. For CV to be successful, a computer must be exposed to many images and videos, analyze them and receive feedback on analyses. Carriers benefit from precise underwriting, reduced need for adjusters to physically inspect properties, reduced human error and fraud, and expedited submissions and claims.
There has been an exponential increase in efficiency of algorithms over recent years. Despite significant advances in machine-intelligence enabled image recognition and analytics, enterprise-scale transformation based on such systems in insurance had proved to be elusive, till recently when insurtechs emerged to capitalize on this opportunity. While some developments such as integrating computer vision into underwriting systems are taking longer, others like damage assessment are changing faster.
Cape Analytics and Betterview are building computer vision solutions for property insurance processes. Using geospatial data, they evaluate building material, roof conditions and square footage, debris surface area, distance from vegetation and myriad other factors that determine the property’s risk profile. CV is enabling this analysis in real-time, at scale, learning from decades of historical data, which is a far cry from sending experts to manually inspect properties.
In auto insurance, a growing number of carriers such as Ageas use CV from Tractable to create car damage assessments and repair estimates. In an earlier post on European insurtech, Tractable was highlighted as a recent unicorn, which across industries, has been a first mover in CV. Tractable uses deep learning for CV, trained on millions of photos of car damage. Ageas policyholders, by submitting accident photos to the CV engine, receive decisions on next steps within minutes — even on an initial phone call. The AI capabilities inherent in Tractable’s CV engine, accelerates Ageas’s speed of response, allowing vehicle damage to be quickly assessed and a full repair estimate generated. The solution supports complex tasks that human assessors have hitherto performed when evaluating vehicle damage. Affected parts of vehicle and extent of damage are quickly assessed to arrive at full estimates of recommended repair, paint and blend costs.
Businesses equipped with comprehensive risk management solutions face lower levels of risk and reduced cost of coverage. Mobileye, that makes advanced driver assistance system (ADAS) for collision avoidance using CV, is one such solution. Mobileye places cameras on vehicle fronts and monitors camera inputs. The system alerts drivers on encountering threats, when drivers veer out of lanes or speed and additionally provides automated braking.
Mobileye effectiveness can be gauged by a Dish Network implementation in their vehicle fleet which resulted in an 88 percent reduction in collisions. Businesses and individuals are adopting this technology to reduce automobile accidents and lower insurance costs.
The use of computer vision in insurance is simplifying underwriting and claims processing, by providing visual assistance to humans in carrying out these procedures. CV is assisting the insurance industry in streamlining operations by curtailing time required and curbing fraud. CV is heralding an era of objectivity and indisputability into insurance claims processes. Carriers gain from detailed perspectives of underlying processes, bolstered by facts and evidence.