Pixurance — Quick Car Insurance
A React Native (Expo) app that lets insurance agents photograph vehicle damage and email an inspection report, powered by a Roboflow-hosted AI model as an inference API.
The Project
Pixurance lets insurance agents photograph vehicle damage and email an inspection report. A fine-tuned Roboflow 3.0 object-detection model classifies the damage, reaching 81.5% mAP and 88.4% precision.
The model was fine-tuned on a self-curated dataset of 6,957 images and 10,768 annotations across 14 vehicle-damage classes, hosted on Roboflow as an inference API, while the React Native (Expo) frontend provides the mobile capture and report workflow.
Technical Highlights
Roboflow 3.0 object-detection model fine-tuned on a self-curated dataset of 6,957 images and 10,768 annotations across 14 vehicle-damage classes.
Roboflow-hosted model served as an inference API, reaching 81.5% mAP and 88.4% precision.
React Native (Expo) for cross-platform mobile (iOS + Android): photograph vehicle damage and email an inspection report.
End-to-end workflow from photo capture to emailed inspection report for insurance agents.
Building a computer vision or mobile AI system?