Stepping Into The Future Of Surveillance: Visual Surveillance Laboratory at NIT Rourkela
Visual Surveillance Laboratory of NIT Rourkela is one of the most cutting edge research laboratories in the institute. Running under the Department of Computer Science and Engineering, this laboratory aims at developing software which is the next step towards Visual Surveillance, Digital Forensics and Biometrics and implementing them in use for societal benefits.
Aiming For The Future
Visual Surveillance Laboratory aims at automating the surveillance system in which the system alerts the operator automatically when unusual activity which is captured and detected by the cameras. This will improve the overall surveillance and result in an improved and more efficient unmanned security system. The laboratory also aims at improving the biometrics system where the recognition will no longer be limited to the retina scan but also face structure, lip and ear placement etc. They also aim at digital forensics in which criminal investigation will be automated through the digital and visual data that the criminal has left.
The researchers who are involved in the laboratory are:
- Prof. Sambit Bakshi
- Prof. Pankaj Kumar Sa
- Prof. Manmath Narayan Sahoo
- Debbrota Paul Chowdhury
- Gautam Kumar
- Tanmay Kumar Behera
- Nayan Behera
Sources of Funding and Resource
Most of the equipment which is being used in the lab is provided by the institute. Some of the cutting edge technologies were granted by NVIDIA Corporation and NITROAA. One of the most powerful equipment in the lab- NVIDIA TITAN V Graphical Processing Unit which is based on NVIDIA Volta Architecture was donated by NVIDIA Corporation to support the Visual Surveillance Lab at NIT Rourkela. To exploit the power of that GPU, a Lenovo P920 workstation was sponsored by NIT Rourkela Overseas Alumni Association(NITROAA). Other than those, it has been funded by different research agencies too.
To get an insight into the Visual Surveillance Laboratory, Team Monday Morning interviewed Dr Sambit Bakshi- one of the professors in charge of the VS Lab. Here is an excerpt from the interview:
Monday Morning(MM): What inspired you to take up this field of research and contribute to it?
Prof Sambit Bakshi(SB): My master's and PhD were centred in this domain. This is a cutting edge technology and many industries, companies and startups are working on this. Along with the various applications it has, there are many technical aspects incorporated and many objectives can be accomplished using this technology and that is the reason why I preferred this field.
MM: How did you get the approval to start this laboratory and research in NIT Rourkela?
SB: I joined back in 2015, my research scholars were then working in this area and I allotted them the projects accordingly. In late 2016, in departmental meetings, there was a discussion on setting up few labs and we formalized this lab as well as 5-6 other labs. At a later stage, we decided on the name and were allocated to the current room.
MM: How did you get the required resources and funding?
SB: Our lab is getting resources from NVIDIA Corporation and the National Institute of Technology Rourkela Overseas Alumni Association (NITROAA) along with that we have tried to get some benefit from our collaborators.
MM: What are the technology and software used?
SB: We use OpenCV, all deep learning technologies primarily. To exploit the power of the GPU(Graphics Processing Unit) we are using various deep learning technologies like Keras.
At the same time, we also work on feature engineering, that is, we would not leave it to deep learning to design the feature which might be good after a long time of training but might not be efficient for a small dataset. But if you can manually see and understand the feature and design it somehow through classic image processing and video processing, that works well.
MM: What societal benefits is this laboratory going to give?
SB: As the name recommends, we work in the area of surveillance and we are working on biometrics also. Technologically the products or intellectual outputs that we generate are being used and can be used in a better version by companies. Science doesn't grow without the policy decisions so we have many technologies developed but to implement them we have policy decisions.
MM: How helpful have been the institution for this lab and projects?
SB: The institute has been helpful since the origin. From providing a space in the Computer Science department to carry on our research works to equipping us with the machines and instruments that we are working on, our institute has a comprehensive role to play. My doctoral scholars are also funded by the government.
MM: What constraints are you facing currently in the project?
SB: There are some minor constraints that we are facing right now in the laboratory. Most of them are technical. There is one major constraint that we are facing is in the setup of a Biometric Studio which will capture the biometrics of a person through a nine-camera setup. For this, we need a big room to set up our studio.
MM: On what basis do you select the scholars working on the project?
SB: BTech students give their preference and we give ours and there is software, that acts as the nearest neighbour matching kind of thing written by automation cell. So automatically BTech and MTech scholars get allocated to us. For doctoral scholars, we take an interview in which we ask some basic ideas of image processing and video processing and second is the basic programming ability that we look for, apart from the general questions that the whole committee asks.
MM: What are the outcomes and results of the lab to date?
SB: A most interesting result that we have recently trying to get is called visual analytics, what happens is that it analyses a video in a very diligent manner and it needs a lot of training.
Let us consider a scenario, on a Sunday at 10 pm, I come to the cabin with a big backpack. You see that I never use a backpack and never come to my cabin at 10 pm. This is a suspicious activity but for a simple camera to detect it suspicious, you need to know my nature of work, my office timings. We want to infix this kind of knowledge in software so that this kind of suspicious activity can be identified. We are trying to train it and also it is difficult to create such kinds of pseudo data otherwise, it is imparting us good results.
Visual Surveillance Laboratory is surely one of those researches that are going to be very relevant in the future. With everything getting automated, security is something that also needs to get developed on these lines. And NIT Rourkela has taken a step into the future by setting up the Visual Surveillance Laboratory.
Team Monday Morning wishes all the best for the future endeavours and success of the initiative.