India's First AI-powered Bicycle Counter
A one-of-a-kind device that helps cities in their push towards sustainable mobility.
This is an Urban Flow project.
As part of the "Complete Streets for Ward 85" project that I've been working on since 2020, we identified the need for a device that could highlight the usage of cycle lanes while also collecting usage data & bicycle traffic patterns.
​
While such devices (bicycle counters) already exist in other countries, none of them are available or suited to Indian conditions. Due to a lack of active enforcement of cycling infrastructure in India, the bicycle counter would need to accurately differentiate between bicycles and other types of vehicles using the cycle lane.
​
As no such product existed, I built the Live Bicycle Counter in 2021-22.
Objectives
-
Make cyclists visible to other road users
-
Encourage cyclists and show that they "count"
-
Build a device that is cost-effective and scalable in the Indian context
-
Collect data & statistics from counters to capture bicycle traffic patterns
-
Enable evidence-backed campaigns for cycling infrastructure
Design & Development
Considering the objectives of the bicycle counter and unique local conditions, it was apparent that an AI & ML based solution would be necessary. While other methods of vehicle counting are common across the world (inductive loops, pneumatic tubes, infrared sensors, etc) these are unable to accurately classify vehicle types and have drawbacks when it comes to scalability, ease of installation and maintenance requirements.
​
Over the course of the project, we created a large & diverse dataset of training images covering various types of vehicles, locations and conditions. This was used to train SOTA (state-of-the-art) AI object detection models for vehicle detection and classification.
​
A rugged & cost effective design was created for the device, which ensured accuracy under diverse conditions, good thermal management and water & dust resistance. The device has retroreflective signage mounted on the front indicating to road users that a daily bicycle count is being displayed.
Technical Features
-
High-visibility display for the bicycle count in real-time
-
Camera with a wide horizontal field-of-view to capture upto three lanes of traffic
-
IoT - 4G LTE connectivity for real-time data upload to the cloud and remote device monitoring & diagnostics
-
< 20W power consumption
-
The ability to count & differentiate between seven types of vehicles:
-
Pedestrians​
-
Bicycles
-
Motorcycles
-
Autorickshaws
-
Cars
-
Buses
-
Trucks
-
Deployment
The Live Bicycle Counter has now been deployed at the following locations in India:
​
-
Bangalore, Karnataka (3 locations)
-
Gurgaon, Haryana (4 locations)
-
Delhi NCR (6 locations)
​
With a simple and secure mounting structure & remote monitoring, it's easy to deploy the counter across various locations in a scalable and cost-effective manner.
Data & Results
Data collected from bicycle counters is critical for cities to make their case towards expanding cycling infrastructure programmes.
​
The Directorate of Urban Land Transport (DULT) in Bangalore, Raahgiri Foundation & GMDA in Gurgaon, and Greenpeace India & Transport for Delhi in Delhi NCR have used various statistics, datasets and visualisations produced as a result of the Live Bicycle Counters for evaluation, consultation and campaigning.