Recently I successfully completed my summer internship as Project Assisstant at Video Analytics Lab, IISc Bangalore. This was my first onsite internship experience which I enjoyed a lot. I got to know about VAL through a senior who had also completed his internship in the same lab. In this post I would be covering the whole procedure of getting an internship at VAL as well as what are the things you get to learn and work upon at VAL. Please note that this post is just about an internship at VAL. Experiences might differ for a Research Assistant or a PhD. Also, this is just a review about the internship experience as a whole and does not include the details of my actual project at VAL as I am not authorised to share the same. ;)
About Video Analytics Lab, IISc
VAL is one of the best known labs currently in the nation working in the field of Computer Vision. Researchers at VAL work over a variety of computer vision tasks ranging from pose estimation to domain adaption and what not. The lab is well known to publish research papers in the best computer vision conferences around the globe.
Intern Application Procedure
If you are thinking of a winter internship at VAL, mail either the professor or one of the PhD scholars of the lab somewhere around start to mid November. Similarly, for a summer internship, mail around mid to end March. Based on availability and your profile, you would get a call for your interview. Interview is aimed to test your basic coding skills and basic knowledge of Deep Learning. In particular, you would be asked to code a few basic problems related to arrays, stacks, queues, dynamic programming, etc. Along with this, a few questions related to basics of Deep learning are also a part of the interview.
Before the start of the interview and after selection, you should make yourself familiar to deep learning frameworks(Tensorflow and PyTorch), ssh, sshfs, jupyter notebooks over ssh, tmux, and whatever project related stuff that you are provided with from your mentor(most likey one of the PhD scholars at VAL). Do mail your mentor asking for resources related to the same before the start of your internship.
Nature of Work
The project that I was assigned to was related to human pose prediction using single RGB image. In ths first week of my internship I got familiar with the general concepts of pose prediction like the different co-ordinate systems used, different joint systems used by different datasets and basics of perspective projection. There are one or two Research Assistants(RAs) also assigned to the same project that you’ll be working on. The RAs are easily approachable and available 24/7 in case you face any difficulty during the course of your internship. Same goes for your mentor. During the next few weeks, I got to proceed with my project where in I was asked to try out different methods/approaches to improve the results of our project. Usually the ideas were proposed by my mentor, and I implemented them and reported the results and improvements to him. Mentors are also open to hear our own ideas and thoughts that we have related to the project.
In the last few days of my internship, I was asked to produce some qualitative results for the project report and try out a few more ideas. I was able to complete most of the work by the end of my internship. I will complete the left out work over the course of next few months whenever my mentor asks me to. There is usually no tighter bound on the time in which we can complete the work although the mentors expect us to complete the work as soon as we can, as research projects require a ot of experimenting and getting the correct visualisations. During the initial stages, getting the correct results might feel exhaustive but with time you get used to it as it is a very important part of your project. As your project is of no use if you can’t convince the readers of your work with the various illustrations available in your report.
Conclusion and Tips
During this internship period, I got an insight on what research actually is and how can one can proceed with a research project. I got to learn how “good” researchers organise and document their work well so that they can contribute something great to the Deep Learning Research community. I consider this internship as a very important part of my career, as it helped me to consider research also as a good option for pursuing in future. I do not like software development that much and in contrary like implementing something new and coming up with new ideas, so I see research specifically in the field of Deep Learning (which I like), as a prominent option for me to take in the future. But who knows what future has to offer. :)