Women and child safety has been an agenda for the government, especially after the Nirbhaya case. In 2019, the Ministry of Home Affairs announced the ‘Safe City Project’ under the Nirbhaya Fund scheme and set aside a budget of around INR 2919.55 crore. 

The project aims to build a safe, secure and empowering environment for women across eight major cities in India—Delhi, Kolkata, Mumbai, Chennai, Bangalore, Hyderabad, Ahmedabad and Lucknow. While the Central government will fully fund the project in Delhi, in the other cities the centre will fund 60% of the project.

“These projects involve identifying hot spots for crimes against women and deploying various components including infrastructure, technology adoption, and capacity building in the community through awareness programmes,” the Ministry said.

Bangalore Safe City Project 

In Bangalore, the safe city project will be implemented in phases, and Honeywell Automation has won the bid to run and maintain the security infrastructure.

Bangalore city police are already equipped with around 1100 CCTV cameras for surveillance. As part of the project, around 7500 cameras will be installed across 3000 locations in the city. Around 6300 of them will be Fixed cameras, 800 pan-tilt-zoom cameras, and 400 high-resolution cameras.

(Source: Twitter)

The tender lists out 2744 locations where the new cameras were to be installed. The number of cameras proposed for each location varies from one to nine, and areas such as critical traffic junctions, market areas, deserted streets, educational institutions, offices and transport hubs are being highlighted.

Among the eight cities the government selected for the Safe City Project, Bangalore was one of the first ones to come up with this Request for Proposal (RFP). “I think Bangalore police saw AI-based technologies be deployed on the ground for women and child safety, how AI can at large help the common citizen on the ground, and make the city of Bangalore more safe and secure. That was the whole motive of this project,” Yatin Kavishwar, Co-Founder and COO, Awiros, said.

Use of AI-based video analytics

In India, law enforcement agencies using video analytics isn’t new. In 2019, UP Police moved beyond CCTV cameras and towards video analytics and analysis of databases to crack down on auto-lifters. However, it could possibly be the first time that video analytics will be used by law enforcement agencies at such a large scale.

“I think Bangalore Safe City Project paved the way for the adoption of video AI in government at large because you require such path-breaking RFP, somebody to take the pole position and take the courage to adopt a new technology, which is just on the horizon.”

“AI has enabled police and administration at large to have a unified view of the scenario across the state. Earlier, what used to happen is that all these solutions used to get deployed in silos; it was at the camera level, not even at the network level,” Yatin Kavishwar said.

Bangalore Police will use the AI-based video analytics service by Awiros for number plate recognition, facial recognition, monitoring areas ridden with issues of crimes against women, health of roads, intrusion detection and crowding, among other things. 

This will enable Bangalore City Police to undertake real-time monitoring and situational analysis and take timely actions to any potential crime being committed by offenders for the safety of the citizens.

Video feeds from all the cameras will be fed to the Data Centre. A Command Control Centre (CCC) will be set up where the videos will be live-streamed. The control centre will use video analytics solutions in emergencies, whether it is to identify a vehicle involved in a crime or a person.

“The platform will obviously enhance the speed of decision-making for the police. I think that is one important aspect of the project,” Yatin Kavishwar said.

Automatic Number Plate Recognition 

The number plate recognition application by Awiros is a highly accurate detection system which can search through thousands of live feeds and track the journey of a vehicle using a couple of attributes (the last two digits of the number plate, for example) and without any human intervention.

“Most of the time, when a crime is committed, a vehicle is involved. So, by using a couple of attributes of the vehicle, Bangalore Police can pinpoint what the track of that vehicle was,” Yatin Kavishwar said.

“The training pipeline is handled by the Annotation, Training and Orchestration Module (ATOM). So, this is our homegrown module, and that data never leaves the platform. What ATOM does is mostly two things. Firstly, it is a data aggregation tool. Now, when these algorithms are running on the platform, a lot of metadata is generated, and most of these are in the form of photos.”

Giving the example of a newly released SUV model, Yatin Kavishwar says that the algorithm might not be trained on the latest models available on the market. “Now, a batch is created of such photos on the ATOM platform. This batch is then pushed to the annotation team. There is an annotation team that sits inside the premise of the customer. Then, the annotated images are used to train algorithms on ATOM which are then pushed back into the Awiros Video AI platform.”

Facial Recognition System

The AI-based video analytics will also help Bangalore Police track down individuals. As per the tender document, facial recognition systems will be used to identify and track down individuals involved in harassing women in public places or involved in any other unlawful activities.

The CCC will pass alerts from the video analytics system to the police on the ground. Simultaneously, the police personnels working on the ground will send the photos of the suspect to the CCC for verification with the criminal database.

However, the use of facial recognition systems by law enforcement has been a matter of concern, especially in the absence of a data protection law in the country. Another concern associated with the use of FRT is that it could be used for surveillance.

“We have been tasked to identify criminals out of a database. Now that database essentially either comes from Bangalore police or any police with whom we are working with. It’s either part of the State Crime Records Bureau database or its part of the National Crime Records database. So for us, we are not profiling individuals; instead, we are profiling the face of the individual against a database of criminals only,” said Yatin Kavishwar.