Crowdsourced Road Damage Collection Service
A public service where anyone can submit photos of road damage without registration; YOLO automatically judges them and plots the results on a map. A memory-efficient design that separates submission intake from detection.
Duration
Short-term development
Team
1-2 people
01 // OVERVIEW
Project Overview
A public service that lets an unrestricted number of people submit photos of road damage for free, with no registration required; YOLO automatically judges the images and plots the damaged locations on a map. Submission intake and image detection are separated into an asynchronous flow, and the detection model is not kept running continuously, which keeps server resource usage low.
02 // KEY FEATURES
Key Features
One-Screen Submission With No Authentication
Just send a photo - no registration required. Location is obtained from the browser's geolocation or the photo's EXIF data. Per-IP rate limiting prevents abuse.
Asynchronous Separation of Intake and Detection
FastAPI accepts submissions instantly, and a worker performs detection via a Redis queue. The model is launched per job rather than kept running, achieving low memory usage.
Map Visualization With Privacy Consideration
Only the detection results and coordinates are retained; the original image is deleted after judgment. Leaflet plots the damaged locations on a map.
03 // TECHNICAL DEEP-DIVE
Technical Highlights
Solution
Resource-Efficient Operation as a Public Service
Rather than keeping the detection model running continuously, the queue-and-worker design smooths out load, and storage is limited to detection results plus coordinates only, keeping the public-facing path lightweight.
04 // IMPACT & RESULTS
Results
No Sign-Up
Anyone can submit
Asynchronous
Intake and detection separated
Low-Memory
Model not kept resident