BSc in Computer Science and Engineering
Bangladesh University of Engineering and Technology (BUET)
Freezing of gait (FOG) is a debilitating symptom of Parkinson's disease that significantly impacts mobility and quality of life. This paper presents FOGSense, a novel detection system using Gramian Angular Field transformations and federated deep learning to identify FOG episodes in real-world settings. Tested on the 'tdcsfog' dataset, FOGSense shows significant improvements in accuracy (10.4%) and F1-score (22.2%) over existing methods, while offering robust performance with missing data and personalized adaptation as symptoms evolve.
A phylogenetic tree represents evolutionary relationships among species. The maximum greedy consensus tree (MGCT) problem seeks a consensus tree with the maximum internal nodes from $k$ conflicting phylogenetic trees and is NP-hard for $k≥3$. This paper presents a heuristic solution with $O(k^3n^5)$ complexity, achieving a consensus tree size of 23.4/26 of a binary tree in experiments. Additionally, the heuristic outperforms random selection in certain tree classes by more effectively handling cluster frequency ties.
This study investigates the social determinants of mental health among undergraduate students in Bangladesh, a developing nation. Over 21 days, 38 students from seven universities used an app to self-monitor their mood, revealing key mental health influencers like academics, family, job/economics, relationships, and religion. The app allowed participants to track mood patterns linked to specific conversations, helping them recognize and adjust behaviors affecting their mental health. Findings highlight unique factors in the Global South, providing insights for culturally tailored mental health interventions.
Women in the global south often seek justice to their online harassment through unveiling the harassers and the screenshots of their sent harassment texts and visual contents before the relevant authorities. Nevertheless, such evidence is often challenged for their authenticity. Our survey (n=91) and interview (n=43) with Bangladeshi online gender harassment victims revealed the depth of the problem, and we set design goals to collect evidence from Facebook Messenger with ensured authenticity. Building on the ‘shame-based model’ of gender justice [12], we designed ‘Unmochon’, a tool that captures authentic evidence and shares with victims’ intended group. Our user-study (n=48) revealed that diminishing authenticity problem may still leave the victim and online gender justice entangled with mob-sentiment, hegemonic legal consciousness, and several privacy aspects. Our findings open up a new discussion on how HCI-design should address online gender justice in such a complex social setting.
Previous Information and Communication Technology and Development (ICTD) research has shown that rural women’s access to digital technologies in many Low and Middle Income Countries (LMICs), including Bangladesh, is challenging because of a number of reasons, including Techno-Phobia. With a view to mediating access for rural women, we designed an one-button communication box, and documented rural women’s user-experiences with it. Drawing upon our discovery of the set of gender specific and culturally contextual problems, this paper also discusses several design and policy interventions to ensure a better Information and Communication Technology (ICT) access for the rural women in Bangladesh.
A novel Freezing of Gait detection system for Parkinson’s disease patients utilizing Gramian Angular Field transformations and federated deep learning with wearable sensors data. The system achieves 86.99% accuracy with multi-channel CNN processing of accelerometer data.
Code for the research paper ‘Unmochon’. Implemented E2E encryption. Full stack development experience.
Thesis project. Contributed a dataset for bioinformatics.
CSE-462 algorithm analysis and presentation
View Project This project focuses on the application of deep learning techniques for automated blood cell classification. This is an object detection project. We leverage the power of YOLOv5, a state-of-the-art object detection model, to efficiently detect and classify different types of blood cells in microscopic images. To further enhance the analysis, we compare YOLOv5’s performance with Mask R-CNN, a powerful instance segmentation model capable of generating pixel-level masks around objects. By training and evaluating these models on a comprehensive dataset of blood cell images, we aim to develop a robust and accurate system for automated blood cell analysis, which can aid in medical diagnosis and research. The dataset used in this project is the BCCD Dataset, which contains 4,888 labeled objects categorized into three distinct classes: RBC (Red Blood Cells): 4,155 objects WBC (White Blood Cells): 372 objects Platelets: 361 objects
The number of personal vehicles usage is increasing manifold. People prefer personal vehicles to commute than depend on public transportation. Finding a parking space in most metropolitan areas, especially during the rush hours, is difficult for drivers. Due to this there is a need to provide sufficient parking places coupled with plenty of slots to help the user park his vehicle safely, also to ensure the user does not end up parking on non-parking area and cause discomfort to pedestrian.Due to this there is a need to provide sufficient parking places coupled with plenty of slots to help the user park his vehicle safely, also to ensure the user does not end up parking on non-parking area and cause discomfort to pedestrian. The idea behind our Android Application- “ParkIn” is to help the user analyse area’s where parking is available and number of slots free in that area.Additionally, the user can pre-book a slot in the area he desires for some consecutive days (along with the daily service) if it is available. This will help reduce the load on the administrator as his physical work reduces drastically and user can search the parking slot through Android Application.User can pay after completion of parking service he received. “ParkIn” Application relieves the user from the hassle of manually searching and waiting for empty slots to park the vehicle.
Introducing page fault in xv6 operating system.
View Project In computer networking, ARP spoofing, ARP cache poisoning, or ARP poison routing, is a technique by which an attacker sends (spoofed) Address Resolution Protocol (ARP) messages onto a local area network. Generally, the aim is to associate the attacker’s MAC address with the IP address of another host, such as the default gateway, causing any traffic meant for that IP address to be sent to the attacker instead. ARP spoofing may allow an attacker to intercept data frames on a network, modify the traffic, or stop all traffic. Often the attack is used as an opening for other attacks, such as denial of service, man in the middle, or session hijacking attacks. The attack can only be used on networks that use ARP, and requires attacker have direct access to the local network segment to be attacked.
Brief Description: Arduino uno Pulse Sensor ATmega32 GSM module (SIM-900a) Bluetooth Module (HC-05) 16x2 LCD monitor Push Button LED
Teaching project. I used this project to teach in youtube and produce online contents. Find the project webpage here.
My first project written in C++ with the help of iGraphics library. Awarded with 2nd runner-up at project showcase competition, CSE Fest, BUET, 2017 Download (exe) and Play
A Heuristic for Maximum Greedy Consensus Tree Problem
Blood Cell Classification using YOLOv5 and Mask R-CNN