safely.ai

​

​

https://safelyai.netlify.app

​

Leveraging AI to improve road safety

Selling your idea

• We build lightweight computer vision models to help municipal corporations and/or city police organisations make better use of their already existing CCTV infrastructure.
• Get in touch with us today to make your city smarter and leverage the power afforded by our fast and accurate models to elevate your city's municipal infrastructure.

What we do

and why we should do it for you

​

• Heavy Vehicle Detection�Detect heavy-vehicles in urban areas automatically. Heavy vehicles are banned during certain times and on certain streets but still roam unwarranted and out of reach of police supervision.

The Problem

Heavy-Vehicle Detection

​

​

• Trucks form the third highest group of vehicles to be involved in road mishaps (12.3%) and road-accident fatalities (15.8%), according to the Road Accidents in India report 2018 published by the Ministry of Road, Transport and Highways.

​

• What’s more, in urban environments, the differences in speed, size and mass between large vehicles and the most vulnerable road users are significant and contribute to both the risk and severity of crashes.

​

Heavy-Vehicle Detection

How we’ll help

• With this model, traffic police departments can automate the tracking of trucks illegally entering truck-restricted road networks.

​

• Our model will also automatically detect the truck's number-plate and store it in a secure database using our custom ANPR algorithm primed to work on Indian roads.

​

• This sort of software does require custom hardware which we believe should be existing in the systems itself. If not, we can assist you through that process as well.

​

Understanding the technology

How we’ll help

​

• Our model detects and tracks objects over time. It uses State-of-the-art methods and is definitely industry-standard
• We store images of potential trucks. The system is made such that more False Positives are likely to be considered than False negatives
• We run our detections through an in-built ANPR system and create reports with license plate numbers, timestamps and locations

​

Seeing the system in action!

Samples

• Timestamp, camera ID, truck photo, and transcribed number plate.

​

• Specialised web-app to generate reports.

• Our model has been custom trained using thousands of truck images and is built on the powerful YOLO model that allows for real-time object detection.

​

​

Day sample - Intense traffic

• Our model, as you can see here work perfectly on Indian roads full of traffic, in fact, it is optimised for them!

​

• Detects vehicles on all sorts of roads.

​

• Works very well on roads full of traffic.

​

Night sample

​

• Detects heavy-vehicles at night successfully.

​

• Works well under shabby conditions.

​

• Some extra lighting would make it extremely robust.

​

Web-app sample

​

• Specialised web app which captures all heavy-vehicles in a video stream

​

• Generate reports easily and maintain permanent records for future reference!

​

Specifications

When and where it will work?

• High quality cameras. If a human can’t see the object nor can the AI!
• GPU systems, ideally already present
• Lighting and all encompassing angles
• Frequent human monitoring

The Implementation Process

​

• Collaborate with traffic engineers to understand hardware and do test runs!
• Adjust the system according to your requirements and the traffic department’s exact specifications
• Pass on system knowledge to technical personnel and intensely assist the implementation process

​

​

Additional Benefits of the system

​

• Can help the department carry out and understand custom analytics regarding truck densities
• We can expand the system to even calculate traffic densities at a checkpoint for further analysis of the traffic flow
• It reduces the human labour and completely automates the system. It is very easy to use compared to the methods employed today

​

​

​

• Vehicle Speed Estimation�Using cameras only, we can develop a system to estimate speed of cars and detect violations. This would require no extra hardware except a regular camera with good vision of traffic flow.

​

• Two wheeler helmet detection�We are also attempting to develop a system another problem: no helmet violation. Using similar technologies we will build software to flag all violations and report the vehicle back to the police

M

FUTURE PROJECTS

Thank you

A project by AImpact

+91 9971793333

arjpan03@gmail.com

https://safelyai.netlify.app

me@nirbhay.co