Projects

A contour-property based Nucleus Segmenting Tool for Real-life cervical cytology images. This is the implementated code of our paper published in BMC Medical Imaging journal on 28th January, 2021 titled:

A contour property based approach to segment nuclei in cervical cytology images 

Tool used: Python(OpenCV, SciPy)

This is the implementation of the website for my paper published in JMIR Medical Informatics on 10th November, 2020 titled:

Toward Preparing a Knowledge Base to Explore Potential Drugs and Biomedical Entities Related to COVID-19: Automated Computational Approach 


Live website available at: CovidBase 

BackEnd tool used: NodeJS(With Express), MongoDB

FrontEnd tool used: JavaScript, JQuery, Bootstrap

This is a website implemented for the doctors in BSMMU to label our segmented Cervical Nucleus dataset. The purpose of the website was to be as simple as possible, with no extra bells and whistle, just logging in and start annotating images. So, it was purposefully made barebones.

BackEnd tool used: NodeJS(With Express), MSSQL

FrontEnd tool used: JavaScript, JQuery, MaterializeCSS

A Rule-based Traffic-scene generation tool. The purpose of this tool is to extract all actor(vehicle, pedestrian) and environment(road,road-type,traffic signal etc.) related information from a Natural Language description and store it as JSON, which can be used to generate still image or a driving simulator(e.g. CARLA) environment/scene. The purpose of these generated scene is to pave way for efficient dataset generation for Autonomous driving tools.

Tool used: Python(Spacy, NLTK, Transformers, UnifiedQA(3B pretrained model), OwlReady2)

Deep Maximum Entropy Markov Model(MEMM) implementation for twitter sentiment analysis.

I used 3 different types of embeddings as features and compare their performance on the Deep MEMM model. These 3 embeddings are:


Tool used: Python(Pytorch, NLTK, Gensim)

F1 Race Predictor tool based on Historical Data. Predict probability of the racers to be in the top 10. 

Also provide a interactive UI allowing users to change certain portions of the model to be trained and selectively select features to be used.


Tool used: Python(Scikit, Flask, Tensorflow)

LAN messenger. Single server hosting multiple(theoretically infinite) clients who can chat with each other and send/receive files over local LAN.


Tool used: Java, JavaFX