36 Hours of Incredible India Police Hackathon 2019

Hackathon — The best 36 hours of my life. In this blog I have shared my experience in an attempt to motivate people for putting efforts in making the world a better place.

India Police Hackathon 2019 is a great initiative taken by Indian Police organized by Karnataka State Police and co-hosted by RV College of Engineering.

Problem Statement

It consists of 5 amazing problems faced by the Indian Police. The exact problem statements are(they are so well framed that I didn't want to rewrite them in my words):

1. Face Recognition
   Find the photo of a missing person, from police database or a string of other official databases or from social media platforms and internet in general. Your task is to develop an app to capture a photo and search for the same on official websites / social media websites / internet using an optimized facial recognition algorithm.
2. MoRTH Data Analysis
   In India itself about eighty thousand people are killed in road crashes every year which is thirteen percent of the total fatality all over the world. Ministry of Road Transport and Highways (MoRTH) monitors and regulates the road transport efficiency system in India, releasing reports of accidents due to various conditions. Road Accident recording and reporting formats have been recently changed. There are 19 reports in all now and the data sets related to accidents with various general parameters will be provided to you.
   What is the interplay of different causes / factors and concomitant impact? Can we find awareness — accident connect? What correlations can be established? Is there any chance of predictions? Can the tool guide the driver in any way? Based on the data points in these reports, develop a Data Analytical Tool (Web App or Mobile) that would help reduce road accidents.
3. Unified Communication Application
   The police department is looking for a user-friendly app to share their thoughts, media, information and ideas just like you do on WhatsApp! It should be a complete official communication tool for the Police. It can later be scaled into different specializations depending on the needs of various internal departments. It can also be interfaced with official databases based on specific access and privileges.
   This being the final objective, the coding fraternity is requested to facilitate creating an app that can share all media forms across all contacts in general and specific requirements demanded. Features like location sharing must be enabled which provides the concept of time and space for Police tasks. And on the lighter side it should take care of any recreational activity or share a joke, the app must be able to do that.
4. Centralised Standard Beat Management Solution
   Each police station’s jurisdiction is divided into beats for the performance of its core policing functions. Beat constable is the main resource performing this task. The presence of the beat constable and the way he performs is at the base of policing.
   The beat system is designed to ensure the same. The gaps still remain humongous. Beat is served at a particular time of the month / day / night depending on the frequency / nature / type of crime. It should take care of surveillance of criminal elements of all types, gather information on the criminal activities in the area and take action in that Beat. What we are looking for is a centralised beat management tool to be assigned to the Station House Officer to ensure better beat monitoring system to reduce crime rate in the respective beats based on beat crime data patterns, cumulatively Police Station in general. Scaling it further, the Circle Police Inspector, the DSP and the SP of the district would be able to utilize this tool for centralised control, monitoring, supervision, strategizing and decision making.
5. Connecting Online Presence for OSINT and Investigation
   Data in the public domain can throw up an Online Presence Matrix of an individual / suspect / criminal / foreign agent, which can provide valuable clues to investigation. The churning of the open source with an ideal mix of technology and domain will throw up what is popularly called OSINT — Open Source Intelligence. From a given input, data, name or photograph, a digital footprint of the person must be must be established by collating all information available on the web. The capability to understand the nature of information available in different open databases, social media platforms and the specific police requirements will facilitate the creation of this tool. In an iterative manner it can be made robust to meet all our requirements.

Submissions

The submission consisted of 3 rounds:
First Round: It was an online Idea Pitch round where the teams had to pitch their solution over the problem statement.
Second Round: It was the offline Prototype demonstration. The teams had to build a prototype in 36 hours and showcase them to the jury of 4 members.
Final Round: Top 10 teams after second round got the opportunity demonstrate their work it to the audience and a panel of 16 Jury members.

Rewards

Along with the cash prize, the winners got National Level Recognition.

My Experience

Our team name was Deep Learners and the problem that we worked on was Face Recognition. The challenge was to develop an accurate system as no one wants innocents to get punished. There were some blockers which made the challenge extremely difficult.

Blockers:

1. Training Dataset of one image per person.
2. Poor Data Quality — Blur Images.
3. Not All images were front facing.
4. Majority of testing data of Unnatural Deaths (includes Suicide, Murders, Accidents) were images where the face was either missing or completely/partially destroyed.

Target Output

Had to successfully match 20 People from a pool of unnatural deaths.

Way to Tackle the problem

I can not share the solution as it was for the government, but I can share the approach. Every team participating in that hackathon started with 0% as none of the state of the art networks are trained over Indian dataset or are made in such a way that they require just an image to train. Our approach was to recognize criminals from criminal dataset and a similarity check from unnatural deaths dataset. We converted the images to latent vector and then clustered all the latent vectors together based on distance. This works for about 50% of the data which is extremely good as the task was to get 20 out of 1000 and we were at 500/1000 and got 97.9 % accuracy in face recognition.

Results

We were India’s top 10 teams. Unfortunately we didn't win.

Learning and Satisfaction

“Many of life’s failures are people who did not realize how close they were to success when they gave up.”

— Thomas Edison (1874–1931)

Winning and losing is a part of life, what really matters is the learning, never give up attitude and the satisfaction. No one is born perfect, everyone starts from zero. After our presentation many teams approached us to find the problem solving approach we used. Some thought about it but did not try and others who tried didn't have the perseverance. One must be willing to try and not quit easily. These hackathons are the places where one can find proper guidance from a mentor or another team. It is a platform that helps you in forming connections with a variety people. This acts as a platform that helps you to interact with people who are ahead as well as behind you. You can be a student and can learn from the people ahead you as they are great source of motivation and information. At the same time, you can act the same for the ones behind you. For me, I made few really good connections, but this wasn’t the best part of the hackathon. It was the feeling of satisfaction that we didn't gave up, it was the feeling of pride that we did something for the nation and the community. The quality work we did and the results we got will definitely be useful for the Police Department. It makes me immensely happy when I think about the impact it will have. Police Department works really hard for the people and I am glad that we could contribute in easing their work.

“If you’re in the luckiest 1% of humanity, you owe it to the rest of humanity to think about the other 99%.”

-Warren Buffett

But this is not where it ends. Currently, many lives are affected by some problems whose solution I can develop. The purpose of this article is to motivate as many people as possible to do something for the society. It can be in any form by developing something for the government, NGOs, NPOs, open sourcing projects, anythings what ever comes to your mind will have a positive impact to the society. I believe that anyone can be in that 1% not by luck but by work.

A huge shout out to my team — Anil, Chirag and Kiran, without whom it would not have been possible.

If anyone has great ideas and wants to collaborate, ping me over my Linkedin id, link is in my bio.