I am Girish and graduated MSc Data Science with AI at University of Exeter, UK. I am a data enthusiast and currently working as an AI Engineer with a Legal Tech Startup. Below is a list of some projects I undertook.
- Created and processed data for 3 different subjects using videos from YouTube and self-shot videos.
- Developed a novel capability of generating and visualising expressions on 3D faces, using deep learning, by creating a robust mapping between valence-arousal space of human emotions and expression parameters of a 3D face model.
- Rendered components of a 3D face to generate facial videos with highest level of realism.
- Used this novel functionality of generating expressions to modify real facial expressions in a video.
- Developed a user friendly and intuitive GUI to enable this facial video editing.
- Undertook a wide set of experiments and user studies to validate the results.
- Accepted for publication at European Conference on Computer Vision (ECCV) 2022, 4th Affective Behavior Analysis In-the-Wild (ABAW) workshop, ECCV 2022, Tel Aviv, Israel.
- Performed facial landmark detection on each frame of the video.
- Cropped face image using these landmarks.
- Detected continuous Valence-Arousal values and discrete emotion labels for each frame.
- Annotated video generation to visualise the discrete emotion labels as well as continuous VA values.
- Checkout the project code and detailed description on Github.
- This project uses the Machine Learning algorithms like K-Nearest Neighbors and implements a deep learning- Convolutional Neural Network with tensorflow using keras API.
- Classifies cell images as infected or not infected.
- The images were labelled as class 0 and 1 signifying uninfected and infected, followed by resizing, shuffling and visualisation.
- Trained and tuned hyperparameters of a KNN classifier using n-fold cross validation.
- Trained and tuned hyperparameters of a Convolutional Neural Network and experimented with different components.
- Performed evaluations on both algorithms, achieving 60% accuracy with KNN and 96% accuracy with CNN on test dataset.
- Checkout the project code and detailed description on Github
- Developed and used Evolutionary Algorithm to optimize and study The Knapsack Problem.
- Followed EA regime of Selection, Crossover, Mutation and replacement of weekest in population.
- Performed a series of experiments and evaluations using different components of EAs like mutation, crossover, population size and tournament selection.
- Checkout the project code and detailed description on Github
- Explored a large dataset of tweets collected from the twitter API during the period March 1st to March 31st 2020
- Sampled and worked with around 100 GB data
- Performed some Basic Stats, analyses and Data lookup
- Created Interactive clustered Maps from geo-tagged tweets with functionality to read tweets by clicking on map markers
- Analysed users and automated accounts
- Identified unusual days and events in UK and Ireland
- Checkout the project code and detailed description on Github
- Coursework under ECMM447 - Social Networks and Text Analysis at University
- This coursework utilises the dataset from Enron communication network
- Performed Network Analysis - Adjacency Matrix, Degree Distribution, Hypothesis Testing - Lognormal Distribution, Assortativity and Disassortativity.
- Performed Community Detection and checked network centrality.
- Performed Compartmental Modelling and simulation of Disease Spread using SI and SIR models
- Checkout the project code and detailed description on Github