Graduate Student
Robotics
Computer Science
Welcome to My Inner Sanctum!
Greetings from the realm of relentless problem-solving and caffeinated brainstorming sessions! You've stumbled onto the kind of busy bee who treats challenges like exciting puzzles, with a determination to assemble each piece into an elegant solution. As a solution-focused enthusiast, I thrive on tackling complex problems head-on, often leading to eureka moments that elicit both high-fives and a few bewildered looks. I’m not just a team player; I'm the teammate who brings both energy and enthusiasm, turning mundane tasks into collaborative adventures.
Everyone likes making money, and that includes me as well. We've all had that moment in our lives where we got really into the stock market and trading but we always wondered how we could make it big. In this project, 4 models have been created that predict the future prices in their own unique way. The models used are LSTM, GRU, 1-Dimensional CNN, and ESN. Different models were used to compare their performances and see which would give better results. The details of this study along with the final results can be found in the report that has been linked below.
Report
GitHub
3-D Structure from motion
Remember those times you tried to capture a 3D object with your camera but ended up with flat 2D images? Well, this project brings those images to life! Using the magic of Python and a collection of Buddha images, we've created a system that reconstructs 3D structures from multiple 2D views. The secret sauce? Bundle adjustment with GTSAM. It's like giving your computer a pair of 3D glasses, allowing it to see the depth and structure hidden in those flat photos. Who knew that a bunch of Buddha pics could lead to such an enlightening 3D experience?
GitHub
GitHub
Report
What happens when you played with an XBOX Kinect as a kid and love Need For Speed? That's right, you end up building a project that uses hand gestures to drive a car. By using deep learning along with ROS integration, the algorithms developed were able to manipulate a two-wheeled car in a Gazebo Simulator. The models used were trained with the help of the 21-keypoints data that were extracted with the help of the Google Mediapipe library as well as the direct RGB images. A comparative analysis was conducted to study the performance of the different models developed.
Do you ever think back to a time when all you had was an old Nokia phone? And what was the one thing everyone did with their Nokia phones? Play the good old Snake game. There was a time when all anyone thought about was getting the highest score possible. Well it might be hard doing it by yourself, but with some reliable code it's more than possible. Here we have a project which uses the concepts of reinforcement learning to beat the game. Models were built using a DQN and a Double-DQN algorithm and their performances compared and analyzed.
Report
GitHub
GitHub
Report
Ever wished your playlist could read your mind? Well, how about your face? This project doesn't just recognize your facial expressions - it acts as your personal DJ! Feeling blue? It's got a playlist for that. On cloud nine? There's a songfor that too. Each facial expression unlocks a specially curated playlist designed to complement or enhance your mood. And guess what? There's even a paper I published on IEEE about this. Now that's something to smile about!
Shredded Document Reconstruction
Have you ever "accidentally" shredded an important document and thought all hope was lost? Well, not anymore! This project is like a digital jigsaw puzzle solver, but for shredded papers. Using the power of Python and some clever Gaussian probabilistic scoring, we've created a system that can piece together those paper strips faster than you can say "Oops!". It's like giving your computer a pair of tweezers and a magnifying glass, allowing it to meticulously reconstruct your documents.
GitHub
Thank you for taking the time to go through my website! If you would like to get in touch, please feel free to reach out via phone or email.
P: +1 (857) 376 1899
E: pranavkn01@gmail.com
© 2023 by Pranav Kizhakkevellat Nair. All rights reserved.