Projects
My self & course projects' repositoriesTry clicking on different tabs to see more projects.
'Repo' button takes you to respective repository.
More projects are coming up!
Machine Reading Comprehension Flagship
Data efficiency comparison of vision transformers to their textual counterparts in extractive Document
Question-Answering (QA) over linear-text novel VQA (Visual QA) dataset, comprising 120k QA pairs.
Dataset:
Graph Neural Networks
Semi-supervised Node and Graph classifications on citation network and protein molecules respectively.
Face-swapping
Face-swapping in videos and images using morphable Basel Face Models, camera projections and GANs.
3D reconstruction
2D images to 3D structures using methods such as affine Structure-from-Motion and COLMAP.
Machine Translation
LLM-based Brazilian Portuguese to English translation: zero-shot vs adapted vs fine-tuned versions of European Portuguese to English model.
Face-swapping
Face-swapping in videos and images using morphable Basel Face Models, camera projections and GANs.
3D reconstruction
2D images to 3D structures using methods such as affine Structure-from-Motion and COLMAP.
Variational Autoencoder
Semi-supervised Node and Graph classifications on citation network and protein molecules respectively.
Iterative closest point
Merge 3D point clouds (coming from different view-based depth maps) to form a single cohesive object.
Single-Shot Detector
Counting objects in image (here it is cigarette packets in vending machine or shelves).
Keyword Extraction
Comparison of unsupervised methods for document keyword extraction: TF-IDF, YAKE, KeyBERT, custom KeyBERT
Machine Reading Comprehension Flagship
Data efficiency comparison of vision transformers to their textual counterparts in extractive Document
Question-Answering (QA) over linear-text novel VQA (Visual QA) dataset, comprising 120k QA pairs.
Dataset:
Machine Translation
LLM-based Brazilian Portuguese to English translation: zero-shot vs adapted vs fine-tuned versions of European Portuguese to English model.
Intent-based Chatbot
A simple chatbot that responds by classifying user intent and finding entities in his/her query.
Elman RNN, LSTM
Classification and auto-regression using Baseline Neural Network, Elman RNN from scratch, Pytorch Elman RNN, Pytorch LSTM.
Graph Neural Networks
Semi-supervised Node and Graph classifications on citation network and protein molecules respectively.
Deep Q-learning
Solving 5-state MDP problem using epsilon-greedy, tabular Q-learning and Deep Q-Network (DQN).
Tic-Tac-Toe
Tic-tac-toe player agent based on MCTS with UCT & Minimax algorithms.
Sudoku solver
Solves any given sudoku using DPLL algorithm and its variants.
Time Series Analysis
Multi-step prediction of climatic descriptors of a city using XGBoost, Prophet, and NeuralProphet.
Bayesian network
A reasoner used to predict required emergency responses near coastal volcanic mountains.
Generalist player agent
2D images to 3D structures using methods such as affine Structure-from-Motion and COLMAP.
Specialist player agent
2D images to 3D structures using methods such as affine Structure-from-Motion and COLMAP.
Neural Network from scratch
Classification of a synthetic dataset and MNIST dataset using scalar and tensor backpropagation respectively.
Interested in my profile?
Contact me for collaboration or assistance with your cutting-edge project or research.
Email Copy Email ID