Projects

TurtleBot3 Patrolling

Date: April, 2024

Overview: In this project, I demonstrate the use of the Navigation 2 framework in ROS 2 through a TurtleBot3 simulation in Gazebo. My main goal is to enable the robot to autonomously patrol an area by navigating through four predefined waypoints on a simulated map. This project serves as a practical introduction to autonomous navigation, where I utilize Behavior Trees for decision-making and showcase both Python and C++ package usage within ROS 2. By leveraging the capabilities of the Gazebo simulator, I test and validate the navigation tasks, providing valuable insights into the fundamentals of robotic navigation in ROS 2.

Implementing Video Swin Transformer Models in TensorFlow 2: A Google Summer of Code 2022 Project

Date: May 2022 - September

Overview: I participated in Google Summer of Code 2022 as a contributor, working under the mentorship of Luiz Gustavo Martins and Sayak Paul. My project focused on implementing TensorFlow 2 versions of the Video Swin Transformer models, converting PyTorch weights to TensorFlow 2, and creating a notebook for fine-tuning these models on the UCF101 dataset. As outcomes, I developed a collection of feature extractors for Video Swin Transformer models on TensorFlow Hub, a fine-tuning notebook with dummy inputs, a conversion notebook for PyTorch weights, and a custom learning rate schedule for cosine decay with warm-up.

GetAudio

Date: July 2023

Overview: In this project, I designed a distributed system to extract audio from videos using a microservice architecture. It involves Docker containers for various tasks, managed as Kubernetes pods. The system uses RabbitMQ for queuing tasks and notifications, with PersistentVolumeClaim for data persistence. Deployed on Azure Kubernetes Service (AKS), it supports automated deployment, replication, scaling, and recovery, ensuring efficient and scalable audio extraction.