One of the best impressions of computer vision came from the movie ‘The Terminator’ when the robot was able to see and identify objects as they moved.
Manual recognition of objects and its counting has been a mammoth workload to people since historical times. Humans have a hard time dealing with fatigue and boredom especially with counting the same object through the day. This monotonous yet very critical job can only be handled by machine for precise performance. With advancement in technology, Computer Vision has emerged as a promising tool to identify and count characteristic things from the wild.
In the world of Insurance where inspections have been quintessential part of the workflow and processes, Computer Vision has the potential to significantly speed up the process, reduce errors, and stop fraudulent reports. The existing system involving manual inspection is greatly affected by the limitations of human beings for consistent and accurate work during these inspections. Computer Vision can help identify objects and avoid false claims/reports. More specifically, the algorithms focus on the automatic extraction, analysis, and understanding of useful information from insurance perspective from a single image or a sequence of images. False claims/reports can happen in any category: Miscalculation, Wrong data, False identification, etc. One of the best uses of Computer Vision is that it could reduce the need for human inspections while providing more accurate and real-time data.
The Actual Problem
One of the main problems faced by Insurance companies is during the time of property valuation through surveys via on-site visits. Deciding whether the property is safe and it can be insured is a time-consuming tedious task for an Insurance inspector. To understand the problem clearly let’s take an example: Consider a Property Insurance where a team of risk inspectors physically visit the place to conduct on-site verification of various risk mitigating objects, checking an efficient arrangement of goods from safety perspective, filling and updating documents and lastly capturing humanly audit able pictures to ensure the building is safe enough to operate. This time and human labor intensive workflow is prone to physical fatigue, erroneous reports, falsification of documents(in case of collusion between inspector and client), fraudulent reimbursement claims, misleading documents provided by property owners, indiscriminate acceptance of property details available from secondary yet unreliable sources. The insurance company needs to solely rely on the Insurance Inspector’s word and the reports.
For premium calculations during property Insurance, Computer Vision has the potential to significantly speed up the process, reduce errors, and lower fraud. One of the top AI companies which is helping these leading Insurance firms in India to address the said problem is Aindra Labs, headquartered in silicon valley of India, Bangalore.
They use state of the art algorithms based on Deep Learning to identify the common objects like Fire Extinguisher, Smoke Detector, Electric Panel, etc (and the list goes on) to verify if the place can be insured with Fire Safety or not. The algorithms identify these specified (by the Insurance Company) key objects from the images clicked using mobile phone and recognizes what type of object is being captured. This smart phone based application can capture data in offline mode and designed to be used for on-the-field property inspection and its valuation report. Subsequently, an automated valuation report based on identification and counting of various subjects-of interest using the captured images on the AI powered visual analytics platform is prepared for the underwriters of Insurance companies.
Object recognition is a well-researched subject and it is still in the process of addressing more versatile challenges. The on-going research has come a long way since its inception however we still need to do much more to deliver to high expectations to hype created around the Artificial Intelligence and Machine learning field. However, With deep learning algorithms becoming more robust and object recognition data sets becoming more and more diverse, the algorithm engineering has vastly improved in recent times.
As the speed of object identification and categorization increases, so does the scope of applicability in the Insurance and other fields.