Skip to content

R3plica

Safer Surveillance With AI

AI, CCTV, Computer Vision4 min read

Mihir Rajput

Meet Mihir Rajput

Computer Vision Engineer @ Wobot Intelligence

Gujarat, India 

Mihir Rajput is a Computer Vision Engineer, Deep learning practitioner, ML, and Data Science enthusiast. He has proficiency in Java spring boot full stack developer and different cloud technologies like AWS, GCP, and Azure.

Mihir's special expertise is in developing, training custom AI models, optimize them, and deploying them in CPUs, GPUs, Cloud servers, and Edge Devices.

He enjoys reading various deep learning research papers and experimenting with different vision-based datasets.


What took you to follow a career in AI?

Since childhood, I was very good at computers and technologies. I have explored and studied different parts of the IT field like Computer Hardware, Networking, Penetration testing/hacking, Software development, etc.

Software development and programming was my favorite among all of the other streams. So I decided to start my career in it. When I was pursuing my B.Tech I made some web applications using PHP as a freelancer. Also, I was very passionate and enthusiastic about new technologies.

So I started to learn Artificial Intelligence and Machine Learning from online courses.

I found that these technologies are at a very pre-mature stage. Additionally, I love research, experimentation, and coming up with new ideas, develop new things. So there were very good opportunities for me here.

One can easily say that AI is future, AI is becoming more and more powerful/popular these days. Everyone is adapting it and making so many useful applications. So with these reasons, I was pretty obvious that I want to work with/in AI.

Favorably just after completing my B.Tech, I got placed in AI and computer vision-based company - Wobot Intelligence.

Define intelligence and the differences with AI?

In my opinion, intelligence is past experience + think/process + action/react/predict. We humans develop our intelligence from experiences.

Of course, there are some more aspects from a biology and synecology point of view but let us keep it AI-oriented. Intelligence is highly dependent on past experiences.

A simple example of this would be a child learns to speak or understand the language. As a child grows, it hears lots of words repeatedly, Its brain stores/processes all the data and tries to understand them well.

After some time it gets able to recognize some words properly. This is a simple example of how a child’s brain learns from lots of data and becomes intelligent for certain languages.

The same rule applies to Artificial Intelligence as well. In AI we make machines or program which learns certain tasks from lots of data, image classification for instance.

The important difference from my point of view is, Artificial Intelligence is not perfectly accurate as human intelligence as of now at least.

Although, AI systems have already outperformed experts in many domains like cancer detection and playing chess. As time goes and as more and more data is produced/collected, the AI system will outperform humans in most tasks.

Explain how AI is implemented on CCTV?

Let’s start with a simple use case where there is an office, it has around 500+ staff members.

So it’s a very tedious and complicated task to manage staff attendance and it’s almost impossible to know which employees are not coming on time obviously until we assign this specific task to humans.

But still, it doesn’t solve the problem what if the supervisor has some friends. The supervisor would probably not be complaining about his friends. Now let’s bring AI into the picture, assume there is a CCTV camera on the gate and it captures gate entry area footage.

Using AI(Facial Recognition) company would easily know which employee is coming at what time. Also, AI can identify latecomers and send alerts to directors or managers directly.

This is a simple example of how AI solves real-world problems using CCTV footage.

Explain how AI could be linked with Computer Vision?

Computer vision is my most favorite domain in AI. Computer Vision highly dependent on deep learning, which is the subpart of machine learning, and also machine learning is the subpart of AI.

Consider CCTV cameras, images, videos these all are vision-based aspects. Whatever applications we make using these data and AI is a computer vision-based AI application.

The most important aspects of computer vision are listed below, and these are widely used in AI robots, drones, self-driving cars, etc. to fulfill a vision or sight capabilities.

Object Classification Object Recognization Object Localization Object Detection Facial Recognition Image Segmentation.

What role models have been your greatest inspiration to get into AI?

Speaking of the role models there are many for me, and most of them are hardcore AI leaders.

I will start with Hadelin de Ponteves and Krill Emerenco from Super Data Science, Hadelin and Krill were the instructor and mentor of my first and many AI courses.

Also, Andrew NG from deeplearning.ai. I have learned lots of stuff from them.

Then the father of AI John McCarthy, Alan Turing, and others who made the AI possible for us.

Then finally actual game changers using AI like Elon musk and Jack ma.

Stephen Hawking and Elon Musk believe AI will change our species? What do you think?

In my opinion, AI is changing the world rapidly. I am sure that it will make human life easier and entertaining but I can’t surely comment about species.

Also if AI would get the ability to change our species, it wouldn’t be in the near future. This kind of evaluation takes decades. So we can’t really predict that kind of ability of AI right now in the 2020s. Because AI is not that strong at this time not even 100% accurate in most scenarios.


For more about Mihir Rajput projects go to @mihir_rajpu