A large corporation in North America operating in Real Estate related services space wanted to improve the quality of home inspections and reduce the time taken by their home inspectors. One opportunity to help them was with the pictures they take as they go around the building during the inspection process. The inspectors would take a picture of the identified problem and later on manually enter text describing it. We proposed a solution that would make it easier for them to enter this description.
The solution was to create a Machine Learning model by giving training data taken from actual images of problems. The model was created using Google VertexAI Vision service and providing custom labels for various problems that needed to be identified. The mobile application that is used by the Inspectors was updated to send the pictures to the model, and the problem identified. A description of the problem would be generated using LLM and it is sent back to the mobile application which can be accepted or modified by the inspector.
The model was initially deployed with a moderate number of frequently seen problems. It was received well by the users as it saved them several minutes of manual entry. More images are being captured and the model is continually updated to be capable of identifying more problems. This solution has enabled the customer to improve the user experience on the mobile application and also improve the quality of the outcome as the system prompts the user with potential problems that they might have missed.