Pramath Krishna

I was involved in an interesting research topic which aims to model human to human interaction through a stream of egocentric life log images as an early social marker for degenerative cognitive impairments like Dementia. The main objective of this project was to develop a robust end-to-end deep learning model that could accurately detect interaction levels between the life logger and the subject in the image. It was stimulating to work with egocentric images (images automatically acquired by wearable cameras) since they have their own set of challenges. For instance, the camera wearer cannot be directly observed in the images therefore the camera wearer’s actions or expressions need to be inferred from the reaction of the subject(s) in the images. 

Being my maiden research project in Artificial Intelligence (AI), I spent the first one to two weeks familiarising myself with the basics of AI by reading widely and clarifying my doubts with other team members. Then, as I got down to apply this knowledge, that’s when things started to pick up fast and I was able to accomplish project milestones in a shorter time frame. Before I knew it, I was fully immersed in all aspects of the project and was involved in the entire research cycle, from literature review to data annotation to data processing for obtaining insights to coding up the models to running experiments for  validating the developed model, and finally putting everything into a research paper (work in progress). It has been an eventful journey thus far.

My typical week starts off with a team meeting every Monday where the other team members and I would update each other of our progress and list out the weekly action items. Then I would spend the rest of the week reading research papers to understand the different approaches explored so far, coding up scripts to process data automatically, developing and testing the model code. My supervisor, Dr Vigneshwaran from the Social & Cognitive Computing (SCC) department would also check in weekly on my progress and guide me towards achieving my tasks for the week. Another significant chunk of my time spent was running experiments to gather data on the model’s performance under different conditions and conducting ablation studies.

Q: How has this internship benefits you?

Besides being given the autonomy to plan my weekly tasks, I have gained a lot from this internship, including my technical skills and knowledge in the areas of computing and AI as well as soft skills like communication and project management. In addition, I now have a greater understanding of how these deep learning models work and was also exposed to using different statistical measures to test the model against industry standard evaluation metrics and benchmarking. During my internship, I was fortunate to have worked closely with a friendly and approachable teammate, Zhang Chen, a researcher engineer from SCC dept, who was always there to answer any questions I had. Moreover, we often had several short discussions every week which allowed for collaborative learning to take place. 

Lastly, a special mention to my supervisor, Dr. Vigneshwaran for giving me this opportunity to be part of the team. This internship has broadened my perspective and exposed me to the computational research world.