"Though, open access and subscription based journal have both got pros and cons, to my mind, it's an open access journal that can make more impact on society and the research community if the quality of publication is maintained."
How did you gain interest in Computer Vision and Pattern Recognition?
While studying in IIT Bombay, I had started investigating MRI Image Processing and later ECG and EEG signal analysis. During this study, I came to know about physical significance of a number of mathematical theories and signal processing algorithms. It was interesting to see the applying theories for various real-life applications. Since, then I had been reading many research papers in the area of computer vision and pattern recognition. Eventually, it became routine to read research papers, implement them, observe the results and apply them in some useful systems or applications. As I investigate more and more in this field, I could experience so many unsolved problems in existing systems and these problems lead me towards future works. While working for my last two projects in Europe, I became fascinated about this field and the work I have been doing. The one and only one question hits me all over the time is "If Human eye can, why can't Machine?".
What trend do you foresee in Computer Vision and Pattern Recognition?
Well, there are many low level problems still unsolved in this area and need to be investigated, but I will talk here about applications that can inspire one to carry out research in this field. First of all, I would like to mention that there are so many vision systems being used in industries, offices, home, public places and secured/important places in the form of biometric systems, surveillance cameras, visual quality testing , video call, detection systems for missing components, etc. However, I see most of these systems are specialized for one or a couple of tasks. To have generalized vision systems, robotics and automation is the area that is inspiring the research community to blend various technologies of hardware and soft intelligence. There are incidences like terrorist attack which is compelling authorities to have electronics surveillance at public and important places. Latest, tsunami in Japan has witnessed the help provided by robots in nuclear plants. Biomedical is the area where signal processing and vision systems are going to play critical roles. Hazardous areas like coal mines can minimize the risk of human casualties by encouraging the use of vision systems. E-governance can take help of systems based on pattern recognition and UIDAI project is a latest example in India where biometric identity will be assigned to each one of 1.2 billion citizens. The real need in this area is to find the techniques that can bring the capabilities in computer or machine similar to the ways human eye sees and interpret! Can we...?
What did you investigate in your last publication?
In our last publication which was based on tracking of human being in thermal imaging, we developed the algorithm to track the human beings using Wigner distribution and pixel association in spatial and temporal direction. We also build a particle filter based on Wigner distribution and evaluated different types of features candidates to be used in particle filter. Still, there are lots to be done in this direction as we have been motivated from the fact that tracking can be simplified significantly; if we use only thermal imaging modality over the visible camera.
How do you approach research? How long does it take to complete a paper?
After identifying a problem, first I decide the way the experiment will be performed with clear understanding of the hypothesis to be verified. The hypothesis can be generated with rigorous literature survey in the area. Of course, previous knowledge and experience add more clarity in objectives and expected outputs of an experiment. The most important phase of experiment, especially in biometrics research or human tracking, is to decide the dataset to be used for evaluation. Having annotation for ground truth or conditions for the available database is always preferred.
The paper preparation to submit in reputed conferences requires around 4 to 6 months. For journal, it may require around 8-12 months. This duration can be reduced if one is dedicated to a single area.
What are the most interesting or surprising things you have learned about since your arrival at your current workplace?
It's always a new learning for me in my current workplace. Working in group adds more information in my knowledge portfolio. When I started creating dataset for my current research project, it was interesting to build GUI to make annotation process semi automated (if not fully automated) instead of doing everything manually. Finally, after creating the dataset, I was surprised to see that our experiment is going to use around 20 thousands files and will take around 2 -3 days to execute the simulations.
What do you consider your most successful moment?
When one of my paper got accepted for OTCBVS, CVPR 10 conference and I attended it in USA. Of course, getting a chance to work with SOCIA Lab members in University of Beira Interior was an asset for my research career.
What would be your recommendation(s) and advise to young researchers who would want to pursue research in Computer Vision and Pattern Recognition?
Do rigorous literature survey and have a "thrust" of publishing papers in conferences and journals. Discussion and brainstorming are very much required for innovation. Well planed and organized processes can make a difference in achieving high quality research.
Were you inspired by a famous scientist or engineer?
Not unfortunately. I think innovation is an integral part of professional life and real inspiration comes from the need of research in today's networked and highly closed world.
What is your feeling about open-access publication? Do you recommend it?
Though, open access and subscription based journal have both got pros and cons, to my mind, it's an open access journal that can make more impact on society and the research community if the quality of publication is maintained. There is a need to have specialized topic/area wise listing in journal web portal. If given a choice, I would prefer open-access journals for my research work over subscription based and recommend same to others.
Playing, travelling and watching movies.
Click here to view all interviews
Would you like to take part in the IJCSI Interview Program (IIP)? Fill the Interview Application Form