So, I have accepted an offer for a PhD position in Sweden. While I was applying for PhD positions, I wrote a couple of Research Proposals along with a Motivation Letter. I have previously written about my Research Proposals here:
Research Proposal 2
Unmanned Aerial Vehicle (UAV) Visual Perception for Safe Landing
This article is about my Motivation Letter which is as follows:
I got introduced to Computer Vision when I took Digital Image Processing in my 5th semester. I took my first step towards Machine Learning when I decided to study Data Mining in my 6th semester.
After completing my sixth semester I worked at ReVeaL Lab as an intern on DiDi Research Algorithm Competition for which I made a model that predicted number of taxis required in a certain area at a certain time on the basis of features like weather conditions, traffic situation in different regions at different time intervals. I took up another course Deep Learning For Perception so that I could have in depth understanding of some advanced topics with practical work.
After Data Mining I had decided to take up my Final Year Project (FYP) in this field to work on real world problems making models for real time systems. I worked on nerve segmentation in a group of three as an FYP and our goal was to identify nerve structure from ultrasound images of the neck for the better placement of pain management catheter.
After graduation, I decided to continue my affiliation with ReVeaL Lab and worked there for about a year as I enjoyed working in a laboratory environment working on interesting research problems some of which include Autonomous Driving in Car Simulations, Image Super
Resolution, Mastering Tic Tac Toe using Self Play and Reinforcement Learning, Blind Textual Deblurring and Neural Steganography. I got an opportunity to work as a Summer Intern at Vision Processing Lab, Information Technology University where I worked on an Indoor Self Driving Car and it was during this time that I decided to continue my studies to enhance my skills and knowledge in the field of my interest and as a result I enrolled as a master’s student at NUCES in 2018.
I have recently given my final thesis presentation on Camera Calibration. My thesis work got accepted at ICASSP 2022. I got the Best Student Paper Award for the same. The code and dataset of my thesis are available on github. I have a published paper Cross-View Image Retrieval-Ground to Aerial Image Retrieval Through Deep Learning on metric learning for image retrieval. I regularly write articles on medium, so far, the topics I have covered include reinforcement learning on games, image compression, camera calibration, depth image generation, web application deployment, spark data frames, turtlebot, model deployment, object detection, Knowledge Graph Completion and Kalman Filter.
I also have some open source implementations to my name including AlphaGo Zero on Tic Tac Toe and ACF Based Region Proposal Extraction for YOLOv3 Network Towards High-Performance Cyclist Detection in High Resolution Images. More details regarding the implementations can be accessed through my github, medium and youtube profiles.
For the last 5 years I have worked on several problems ranging from object detection, neural steganography, nerve segmentation, autonomous driving, humidity and posture detection, driver prediction, language translation, image retrieval, knowledge graph completion and distraction detection. I believe that this position will also provide me an opportunity to gain exposure to new technologies and problem areas and impact real lives on a much larger scale than I am currently doing through my writing, demos and codes.
During my PhD, I want to work on revolutionary ideas with an impact similar to Mastering the game of Go without human knowledge, Distinctive Image Features from Scale-Invariant Key-points, A New Approach to Linear Filtering and Prediction Problems, Histograms of Oriented Gradients for Human Detection, Imagenet classification with deep convolutional neural networks etc. Specifically, as far as scene understanding is concerned, I believe that an end-to-end model similar to AlphaZero is the solution for applications like autonomous driving moving forward with sub-tasks including Localization, Detection, Segmentation etc. and a possible direction could be to combine Reinforcement Learning with Student Teacher networks which I would love to explore in further detail during my PhD.
That’s it for now. See you later.