Photo Widget
We developed a fully guided photo capture experience powered by in-browser AI, helping users take accurate car photos from all required angles, and enabling instant damage detection—all without server-side inference.
01 The Challenge
Insurance and car marketplace platforms required high-quality, standardized photos from specific angles — but users consistently submitted blurry, poorly framed, or incomplete photo sets, leading to claim rejections and re-submission loops.
02 Our Solution
We built a browser-native guided photo capture widget using TensorFlow.js for on-device inference. The widget provides real-time guidance overlays showing correct framing, detects when the angle matches requirements, checks for blur and lighting, and flags visible damage — all running at 30fps in the browser without any server calls.
03 Tech Stack
04 Results
Photo re-submission rate dropped from 45% to 6%. The widget runs at 60fps on mid-range mobile devices. Zero server infrastructure required for inference — reducing operational costs to near zero.