Systems and methods for automated property damage estimations and detection based on image analysis and neural network training
US11935219B1 · kind B1 · utility
Assignee
Inventors
Key dates
| Filing date | Apr 8, 2021 |
| Grant date | Mar 19, 2024 |
| Priority date | — |
| Expiry date | Jun 1, 2042 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06V2201/08
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
Intelligent prediction systems and methods of use to train a neural network model to analyze images of property damage to detect and predict property damage of a property, the neural network model during training configured to (1) switch between one or more synthetic images comprising pixel-based masked annotations of damaged property from a synthetic engine and one or more real images comprising bounding box annotations of damaged property from a real database, and (2) freeze inactive class training to prevent learning on one or more inactive classes comprising one or more pre-determined missing annotated labels in the one or more synthetic images and/or the one or more real images.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.