• Why we seem to lose interest in things over time?

  The philosophy behind some of my research ideas with a gentle introduction to submodularity - something I worked on during my time with the research group at AitoeLabs and the Indian Institute of Technology, Bombay.




• How might we make people read 1000 books in their lifetimes?

  Some thoughts about reading, possibly helping people read more using "AI", and my time at the Digital Product School in Munich, Germany.




• Are you disciplined if someone has to supervise you?

  The great J. Cole says "...the only real change comes from inside". Why you're probably overstating your capacity to learn if the only signal you get is external.




• Thoughts on surveillance and computer vision's part in it

  We like to have information. We like to have information over others. Why having the flow of information set up as a zero sum game is a problem.




• Learning through Recreational research

  Recreating AlexNet from scratch in (old) Keras and understanding the math behind convolutions and pooling from first principles.




• Working with Less Data - Image classification with very few images.

  Using Kaggle's Dogs v/s Cats to advocate the use of data augmentation and pseudo-labeling on pretty much any image classification tasks.




• Is your dog cute or not?!

  Using state of the art AI find out if your dog is cute or not!




• Kaggle - Dog Breed Classification

  My 527th ranked submission to the dog breed classification problem using a machine learning framework I wrote and the lessons learnt.




• Analyzing flags of the world

  What would the combined flag of the world look like?

Published Papers/Longer Reads

• Improving Visual Recognition using Ambient Sound for Supervision

• Learning From Less Data: A Unified Data Subset Selection and Active Learning Framework for Computer Vision

• Demystifying Multi-Faceted Video Summarization: Tradeoff Between Diversity, Representation, Coverage and Importance

• Vis-DSS: An Open-Source toolkit for Visual Data Selection and Summarization

• Deployment of customized deep learning based video analytics on surveillance cameras