Debajyoti Chakraborty

Debajyoti Chakraborty

Graduate Student @ NEU

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 1

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

projects

Phototron

Indian Institute of Technology (IIT) Guwahati - Techniche

  • September 2016 - February 2016 : Blanka Botz
    • To make an autonomous robot that can track a colored ball while it is falling through an incline and successfully catch it.
    • Extensive use of OpenCV for Python with special attention paid to color space understanding, contour detection, and centroid analysis.

Quora Question Pairs

TCS ExOP for Artificial Intelligence - Enterprise Intelligent Automation

  • June 2017 - April 2017 : Researcher
    • Development of an autonomous system that identifies question pairs that have similar intent and clusters them into sets from the same subject or mark them as duplicate questions, thus improving overall querying time and reducing the need to write redundant answers.
    • A semantic analysis of the vectors was performed to find the probability of them being the same using a Siamese Manhattan LSTM architecture with an accuracy of 96.16%.

SBI Automate for Bank

State Bank of India (SBI) - Automate for Bank

  • August 2018 - January 2018 : Computer Vision Researcher
    • Design of a secure mobile banking application for State Bank of India that includes enhanced login via face recognition, a feature to scan and submit mandate forms/cheques with automatic signature matching and verification.
    • Login was supported by a simple CNN tuned to detect faces in the wild and then compare the face embedding with the ones stored in the bank database for similarity.
    • Document processing was achieved through a smart cropping system with the help of OpenCV correcting for geometric distortions or deformations and edge/contour detection.
    • Same was followed by localization of regions of interest in the document and extracting essential information like the MICR code, amount, payee details and signature.
    • A convolutional Siamese Network provided the best results when tested for signature comparison and verification for features like dimension ratio and signature density.

Traffic Sign Detection

CS 703C: Artificial Intelligence - Netaji Subhash Engineering College

  • April 2018 - February 2018 : Undergraduate student
    • Can detect and recognize all mandatory traffic signs and most cautionary signs in video sequences recorded by an onboard vehicle camera using the OpenCV cascade of classifiers.
    • Solution was implemented for use on a daily basis which aided a lot of people using it, who often became confused with the minuscule changes of signs all over a diverse country such as India, particularly tending to the elderly and specially-abled, easing their driving experience.
    • A vehicle speed checking mechanism was also added later on to the system which dramatically reduced if not prevented speed violations, for streets where the speed limit information was not available digitally, such as on Google Maps.

Semantic Indexed Document Search Engine

Tata Consultancy Services (TCS)

  • March 2019 - November 2011 : Software Engineer
    • Implementation of a document search engine using an integration of full-text and much more efficient, semantic search, indexed by relative closeness to the query term.
    • Sentence and document-level embeddings were developed from the ground up with ideas referencing word embeddings for a better understanding of relative content.
    • Search terms were corrected for query auto-completion, filtering, and augmentation which after being fed into the engine returned accurate ranked query results.

Eternal Labyrinths

CS 5010: Program Design Paradigm - Northeastern University

  • December 2021 - September 2021 : Graduate student
    • Created a graphical adventure game in Java Swing where users can navigate procedurally generated dungeons, tackle puzzles, amass rewards, and vanquish formidable monsters to reach goal position.

Playyn - Online Music Store

CS 5200: Database Management Systems - Northeastern University

  • April 2022 - January 2022 : Graduate student
    • Built an online music store in Python, MySQL, and Flask, catering to an individual’s specific audio interests, by showing a detailed audio analysis and feature breakdown for tracks uploaded by artists and bands.

publications

Multimodal Segmentation of Brain Tumours in Volumetric MRI Scans of the Brain Using Time-Distributed U-Net

Abstract

Brain tumour segmentation poses a challenging task even in the eyes of a trained medical practitioner. Traditional machine learning algorithms require hand-coding features from images before they can learn to identify the regions. Deep learning can solve the problem of detecting tumours with precision and even segment it. Neural networks can learn a hierarchical representation of features from the data by itself. We use a time-distributed architecture for U-Net based deep convolutional neural networks (TD-UNET). We tested our network against the MICCAI BRATS 2015 dataset that comprised 220 high-graded gliomas (HGG) and 54 low-graded gliomas (LGG) and yielded a test case accuracy of 58.3%.

Dutta, J., Chakraborty, D., Mondal, D. (2020). Multimodal Segmentation of Brain Tumours in Volumetric MRI Scans of the Brain Using Time-Distributed U-Net. In: Das, A., Nayak, J., Naik, B., Pati, S., Pelusi, D. (eds) Computational Intelligence in Pattern Recognition. Advances in Intelligent Systems and Computing, vol 999. Springer, Singapore. https://doi.org/10.1007/978-981-13-9042-5_62

Exploring reinforcement learning techniques for discrete and continuous control tasks in the MuJoCo environment

Abstract

We leverage the fast physics simulator, MuJoCo to run tasks in a continuous control environment and reveal details like the observation space, action space, rewards, etc. for each task. We benchmark value-based methods for continuous control by comparing Q-learning and SARSA through a discretization approach, and using them as baselines, progressively moving into one of the state-of-the-art deep policy gradient method DDPG. Over a large number of episodes, Qlearning outscored SARSA, but DDPG outperformed both in a small number of episodes. Lastly, we also fine-tuned the model hyper-parameters expecting to squeeze more performance but using lesser time and resources. We anticipated that the new design for DDPG would vastly improve performance, yet after only a few episodes, we were able to achieve decent average rewards. We expect to improve the performance provided adequate time and computational resources.

Vaddadi Sai Rahul & Chakraborty, D. (2023). Exploring reinforcement learning techniques for discrete and continuous control tasks in the MuJoCo environment. arXiv preprint. arXiv:2307.11166.

Screening Mammography Breast Cancer Detection

Abstract

Breast cancer is a leading cause of cancer-related deaths, but current programs are expensive and prone to false positives, leading to unnecessary follow-up and patient anxiety. This paper proposes a solution to automated breast cancer detection, to improve the efficiency and accuracy of screening programs. Different methodologies were tested against the RSNA dataset of radiographic breast images of roughly 20,000 female patients and yielded an average validation case pF1 score of 0.56 across methods.

Chakraborty, D. (2023). Screening Mammography Breast Cancer Detection. arXiv preprint. arXiv:2307.11274.

talks

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.