Potholes, debris, glass, blocked bike lanes, bad surfaces, and close-passing cars are dangers you often only register the moment they hurt you.
02 / The city
The city's blind spot
Planners have almost no granular, ground-truth data on where cycling is dangerous. Streets get fixed based on crashes and complaints — not the near-misses riders face every day.
Most cycling danger isn't at famous intersections — it's on everyday roads where cities lack real-time hazard data. Blind Spot generates that data automatically.
02The evidence
The numbers cities can't ignore.
Cycling danger is concentrated, measurable, and chronically under-reported. Better data leads directly to safer roads, more riders, and lower emissions.
~1,000
U.S. cyclists killed each year
In motor-vehicle crashes
120K
Cyclist injury ER visits annually
Crash-related, U.S.
59%
Of cyclist deaths away from intersections
On ordinary road segments
3×
Road-surface crashes hit cyclists more
Vs. other road users
91%
Fewer bike injury crashes
On one protected-lane corridor
+18%
More cycling after protected lanes
Across 72M bike-share trips
Where danger lives
Danger isn't at intersections
Most serious cycling crashes happen on everyday road segments, and nearly a third involve overtaking vehicles — exactly the near-misses cities never see.
The data gap
Cities lack the data, not the roads
The U.S. DOT created bicycle-safety performance measures because agencies need better cyclist data. Blind Spot generates that ground-truth data automatically.
Why now
Crowd-sourced data actually works
Recent research shows crowdsourced cycling data can accurately model citywide cycling activity even when official counts are limited — validating the whole model.
Blind Spot turns every bicycle into a mobile infrastructure sensor — giving cities the real-time safety and accessibility data they need to make cycling safer, increase ridership, and accelerate sustainable transportation.
03Product pillars
Four pillars, one flywheel.
The handlebar hardware is just the capture layer. The real product is the collective map that turns thousands of rides into something a city could never see alone.
01
Effortless capture
The bike notices danger so the rider doesn't have to. A single button and a forward-facing camera turn every ride into structured, geo-tagged data — no extra sensors required.
One-press manual flag
Camera-based fall detection
Continuous GPS ride tracking
Local buffer + sync, works fully offline
02
A camera that thinks
Computer vision turns raw frames into labeled hazards automatically. The map fills itself, not just from manual reports.
YOLOv8n pothole detection
Multi-class: debris, glass, water, blocked lanes
Bike-lane + surface roughness scoring
Face & plate blurring for privacy
03
A self-cleaning map
Crowd data that stays fresh and trustworthy. Hazards confirm when riders re-flag them and expire when riders pass clean.
Live hazard map + density heatmap
Community confirm / auto-expire logic
Accessibility layer for curb cuts & ramps
GeoJSON / CSV export for cities
04
A safety payoff
Safer routing and ride insights make contributing worth it for the individual rider, not just the city.
Safety-first routing around hazards
Personal hazard alerts on your routes
Auto ride recap card + rating
Surface & bike-lane quality scoring
04The real product
A collective map that fills — and cleans — itself.
Every flag and auto-detection aggregates onto one city map: a hazard density heatmap, tappable pins with photo, type and severity, plus an accessibility layer for curb cuts, ramps and surface quality. Hazards confirm when riders re-flag them and auto-expire when they pass clean — solving the stale-data problem that kills crowd maps.
Severe
Moderate
Minor / caution
Safe / smooth
# north star · hazards mapped per active rider per week
05The rider payoff
Routing that keeps you off the bad roads.
A tangible payoff makes contributing worth it for the individual — not just the city. The same crowd data that informs planners routes you around the hazards other riders already found.
01
Hazards ahead, flagged
Confirmed potholes, debris, and blocked lanes on your route surface as alerts before you reach them.
02
Routed around danger
Safety-first routing weighs the live hazard and surface-quality map, not just the fastest line.
03
Ride recap + rating
Every ride returns a recap card scoring the roads you took, so contributing pays you back.
06How it works
From handlebar to city hall.
01
Capture layer
The bike
Raspberry Pi brain with Pi camera, push button, GPS, buzzer and LED. The camera handles both hazard and fall detection. Buffers locally in SQLite, syncs over WiFi or phone hotspot.
02
Processing
Backend
FastAPI ingest, a YOLOv8n detection worker on captured frames, a privacy worker that blurs faces and plates, plus aggregation, routing and export APIs.
03
Storage
Supabase
Postgres + Auth for rides, events, hazards and confirmations. Object storage for photos and recap cards. Realtime so hazards appear live.
04
Consumption
Web & mobile app
Next.js + MapLibre map with hazard, accessibility and heatmap layers, ride history, recap cards, confirm UI and safety-first routing.
Data flow · one ride
01Rider moves → device logs GPS track
02Camera detects hazard or fall → auto-capture frame + event
