Move the map so the pin sits exactly on the barrier. Zoom in for accuracy.
Vadodara Access Map
An open map of physical-mobility barriers in Vadodara and its outskirts.
Broken or missing ramps, high kerbs, blocked footpaths and steps with no alternative make it hard or impossible for wheelchair users, older people and parents with prams to get around. This map records those barriers street by street, with a photo and an exact location, so anyone can check them and the city can fix them.
How to use it
Tap a pin to see the photo and details. Square pins are Om's photos.
Small dots are the AI street scan: green = footpath seen, red = no footpath seen on a built-up street, purple ring = zebra crossing to check for a ramp. Tap one to see the photo and say whether the AI is right.
Report a barrier: pick the type, place the pin, answer a few short questions.
Layers & filters: satellite view, street-level photos, OpenStreetMap gaps, and filters by type and severity.
Roadmap
Now: Om's photos, an AI street scan of every Mapillary photo in the city, and people checking the AI.
Next: shared storage so everyone sees each other's reports. Verified kerb and tactile-paving data added to OpenStreetMap, where every map app can use it.
Later: more street photos (anyone can add them with the free Mapillary app), so the AI scan covers footpaths, not just main roads. Then a ramp/obstacle detector trained on the photos people have checked.
Local news has reported for years that the city is hard to get around for people with disabilities:
"Vadodara's buildings are not disabled-friendly", Times of India, 24 July 2014. RTI replies showed that most buildings in the district lacked ramps, accessible lifts, parking, toilets and signage.
"City yet to adapt to needs of the disabled", Times of India, 4 July 2014. Apart from the collector's office, public buildings and places remained difficult to access, despite repeated petitions.
Those reports are years old. This project asks: what is it like today, street by street?
The approach follows the University of Alberta's Atlas / Click&Push project. Here the AI step uses Mapillary's computer vision on street photos, and people check its answers. The question design follows OpenStreetMap accessibility tagging, as used by MapComplete OnWheels. This is a small, student-led project for Vadodara.
Om's photos: taken in Vadodara and placed by hand from the location Om gave. Type, severity and notes are Om's. Om could not do a systematic walk, so there is no count of places checked.
AI street scan: Mapillary runs semantic segmentation on every uploaded photo (it labels road, footpath, kerb, building, crossing and more). tools/ai_scan.js asks of each daytime photo: is there a road, and is a footpath visible? Built-up streets with none are red. Night photos and photos with no road are not judged. Anyone can re-run it as new photos are added.
Checking the AI: in a spot check of 12 red dots, 8 clearly had no footpath and 4 were unclear. The AI almost never detects kerb ramps, so crossings are marked "check for a ramp", not "no ramp".
AI-suggested barriers (purple diamonds): street photos where a detector, trained on the confirmed labels, proposes a barrier type. Unverified. Check them and it learns.
Severity: High = a wheelchair user cannot pass. Medium = possible but difficult or unsafe. Low = inconvenient.
OpenStreetMap-ready answers: where a standard tag exists (kerb, tactile_paving, wheelchair, lit), the answer is stored as that tag. Verified points can then be added to OpenStreetMap.
Street-level photos come from Mapillary. They are context until someone marks a barrier in one.
OSM gaps are mapped footways and crossings with no kerb, wheelchair or tactile-paving tags. That means unknown, not confirmed bad.
Reports you add are sent for review and appear for everyone only once approved. A copy stays in your browser.
Very few footways in this area are mapped in OpenStreetMap (26 footway ways when checked, none with accessibility tags).
Satellite imagery cannot show the condition of a ramp. It is context only.
Phone GPS can be off by 5–15 m. Om's photos cover a few places, not the whole city.
The AI street scan only sees where Mapillary photos exist: mostly main roads and highways, taken from vehicles, many from 2014–15. A red dot means no footpath was seen in that photo, not that none exists.