Freshers’ 2020: Research at IIIT

One of the best ways I’ve heard someone describe research is this quote by Zora Hurston:

“Research is formalised curiosity. It is poking and prying with purpose.” 

We think about new ways to approach a problem, a new way to formulate a problem, experiment with things that have already been done in another domain and expand it to new ones. While undergraduate research happens in many places, in IIIT, it’s more structured and allows you to start actively contributing to the field early on.

A lot of cool stuff happens in the IIIT labs. Here is a high-level overview of some of the projects in the various labs:

LipSync: Prof Jawahar and some of the students in CVIT developed something they call “Face to Face Machine Translation”. In face to face machine translation, given a video, you not only translate the audio or the voice in the clip but also ensure there’s “visual translation” as well. So they modify the facial features like lip and movement to sync up with the translated audio, This provides a much more immersive experience and has a variety of applications be it in dubbing movies or in automatic translation of TV interviews etc.

Autonomous Cars and UAVs: The robotic research centre’s forte is arguably robotic vision. To that end, they actively maintain a working autonomous car at IIIT, fitted with various sensors and cameras. Professor Madhav Krishna’s team at RRC have been working on this for quite a while and you can occasionally see it driving on campus. They also work on unmanned aerial vehicles, developing cutting edge algorithms for navigation and path planning among other things. 

Automated SciBlogging: Prof Vasudev and his students in the IRE lab created a system to generate blog titles given scientific research papers automatically. Science papers are usually very technical and challenging to understand to the layperson. But with blogs, one can get the gist of the paper, and they play a significant role in making scientific research accessible to a broader audience. Thus an automatic way to generate simple titles from complex sentences can be an essential milestone for future work.