Clear Capital recently announced ClearPhoto, a powerful set of AI-driven rules built into ClearCollateral Review, automating the review of property photos and ensuring they are aligned with the appraisal data and sketch. The technology was developed by Clear Capital’s data science team to further streamline the appraisal process and equip lenders with technology that powers low-risk, high-quality decisions, according to a company release.

At a time when the cost of underwriting has greatly increased because of labor shortages and overall demand, lenders are looking for solutions to use human expertise in a more targeted way. ClearPhoto uses computer vision technology to highlight the right files automatically so underwriters can make more efficient and informed decisions, the company said.

ClearCollateral Review is now enhanced with automated critical checks throughout the appraisal review process, including:

• Confirmation that each photo is correctly labeled.
• Validation that the number of unique rooms shown in imagery matches the appraisal data for bedrooms, bathrooms, and kitchens.
• Validation that the number of bedrooms, bathrooms, and kitchens in the sketch matches the appraisal data.

“One of the last remaining, untamed areas of the appraisal process is the understanding of photo quality, and most lenders and AMCs are manually reviewing photos with inconsistent, slow, and wildly expensive processes,” Jeff Allen, executive vice president of innovation labs at Clear Capital, said in the release. “Employing computer vision and machine learning help streamline the inspection process, lower the cost of collateral underwriting, and reduce the time it takes to get collateral approval. The technology is part of our broader efforts to digitize appraisals and empower the industry to make better underwriting decisions.”

Erik Allen, Clear Capital’s chief data scientist, said the company’s existing platform, capabilities, and unique set of data made the decision to pursue the development of photo artificial intelligence and automated review a natural next step.

“Our proprietary technology is currently making nearly one million automated underwriting decisions per month,” he said. “Our photo review suite is made up of four machine learning models that can review and analyze thousands of photos in our dataset quickly and accurately. The technology has the ability to provide important context in photos, which helps achieve higher-quality appraisals and better underwriting decisions.”