According to the Insurance Institute for Highway Safety, proper use of a bicycle helmet reduces the odds of a head injury by 50%, while the odds of damage to the face or neck are dropped by 33%. While this fact is generally known, it is still hard to estimate the actual number of cyclists that wear helmets. Abto Software R&D team has developed a computer vision algorithm that does just that and allows to analyze whether the cyclists on the video wear number one safety gear.

## Computer Vision-based Cyclists Detection

### Approach I. People Detection & Bicycle Detection

Our first approach employs computer vision-based object detection techniques to detect people and bicycles separately.

### Approach II. Cyclists Detection

To perform cyclists detection we trained a new object recognition model that reached the accuracy of 97% and is visualized by the next video.

## Camera-based Bicycle Helmet Detection

In order to extend our cyclist detection model, we decided to add the feature of analyzing whether the person wears a bike helmet.

## Technologies Used for Cyclists Detection & Helmet Use Analysis

• OpenCV
• Python
• Video processing
• Transfer learning
• Convolutional neural networks (CNNs)

## Areas of Application of Abto Cyclists Detection & Helmet Use Analysis Technology

Abto cyclists detection & helmet use analysis technology facilitates:

• Estimating bike-lane demand
• Planning and development of cycling infrastructure
• Ensuring urban safety and estimating cycling injury risks
• Gathering real-time data on cycling flows

## What’s next

The developed cyclists detection and helmet use analysis technology is the first step to the comprehensive cycling flow analysis solution. We plan to extend our algorithm with cyclists tracking and counting as well as age and gender analysis feature.

###### 1. Insurance Institute for Highway Safety, Highway Loss Data Institute. Fatality Facts 2017. Bicyclists

$${}$$