About

I’m a computer vision enthusiast with research interests in 2D and 3D object tracking and detection, image-segmentation and AR/VR.

Project highlights and overview

1. Image Segmentation from satellite imagery Jan 2017 - July 2019

   The objective of the project was to extract map features from satellite imagery, with a primary focus on extracting building footprints. The goal was to achieve a model that could perform instance segmentation, separating building boundaries, and be sufficiently generalized to identify both small and large buildings.
   
   Dual-scale training was used to learn the representation of both small and large buildings. This setup was tried on the Inria dataset and an Intersection over Union (IOU) score of 76.98 (“the score reported is the result of the submission made in 2019”).

2. Computer vision assisted real-time map making May 2019 - Jan 2020

   The aim of the project was to create an efficient application that could link existing map-making tools with computer vision-trained models. The application aimed to automate as much of the mapping process as possible to streamline the map-making process.

3. Object detection and tracking in changing weather conditions Oct 2022 - Apr 2023

   The aim of the project was to address the challenges faced by computer vision systems in detecting and tracking objects under diverse weather conditions, specifically focusing on tracking cars from a stationary camera mounted on the road. The primary objective was to evaluate the impact of various weather scenarios on existing detection-tracking architectures.
    
   Project Highlights:

   1. Exploring the difficulties associated with accurately detecting and tracking objects in changing weather conditions.
   2. Evaluating the performance of a selected tracking by detection architecture under different weather scenarios.
   3. Identifying limitations of model that performs well in clear weather but struggle in severe weather conditions.
   4. Conducting a comparative analysis between selected tracking by detection based and joint detection-tracking approaches.

Awards & Honors

• Most innovative project at TomTom innovation Day, 2018
• Winner of the TomTom Developers Day - for providing a solution on tackling the problem of fake news on social media, 2018
• Winner of the TomTom 24 hour Hackathon - for building a smart city guide mobile app, 2018
• Winner of the TomTom 24 hour Hackathon, 2017

Updates

[2022-2023] Joined as a student researcher at the Institute of Computer Graphics and Vision
[2021-present] Joined as a Software Researcher at Joanneum Research
[2021-present] Started my masters at Technische Universität Graz
[2017-2021] Joined R&D team, TomTom, India