AI for home inspection

The objective of this exercise is to explore AI and solve one of the real life solution for home inspection.

Please check with your own roof images

The roof is one of the most important features of a house. Not only is it a key support structure, but it also acts as the building’s first line of defense from the outdoor elements such as rain, snow, wind, and extreme temperatures. The common roof problems found in home inspections include a failing structure (sagging), poor installation or workmanship, leaks, missing/damaged shingles, improperly installed flashings, and clogged gutters and downspouts.

When we request for home inspection, the inspector will typically collect all dwelling information and will take lots of property photos. Based on data and photos collected, inspector will prepare detailed dwelling report. ?The data and photos are verified by QA as quality check and photos are segregated.?If issues are observed in roof or shingles then it’s highlighted in report as “roof damage”. ?Such issues may influence underwriting decision and replacement cost.

AI can improve efficiency and accuracy of report by identifying issues like “roof damage”. I am using Google Teachable Machine for training my model. I am working with Image project with Standard Image Model. I collected lots of roof images. I classified my model into two classes “Good” & “Bad”. Good indicates roof is in excellent condition. “Bad” indicates roof is not in good condition. I tuned my model further by providing more images of different dimension, variety & type. ?And then checked with test data, to my surprise the results are outstanding. Happy to see AI solving the real issue.

Please check with your own roof images

Typical home inspection has 20 to 30 photos, if AI can help identify the issues, it will save lots of QA time. It will also improve accuracy of the report and will benefit underwriting decisions.