Antu Roy Chowdhury
Electronics & Telecommunication Engineering student at RUET specializing in practical machine learning, image processing, and full-stack prototyping. I build systems that bridge theoretical research with real-world utility.
5+
Projects listed
10
Experience entries
4
Skill groups
10
Recognitions & awards
10
Research Works
10
Total Citations
2
h-index
8.3
Research Interest
Strength
Applying deep learning to medical image analysis, microscopic classification, and clinical automation. Researching CNN architectures, feature extraction, and model interpretability using datasets like Raabin-WBC and PBC to build reliable, AI-assisted diagnostic systems.
Strength
Designing sensor-integrated smart systems that bridge embedded hardware, wireless communication, and real-time monitoring. Focused on autonomous systems, sensor fusion, and prototyping scalable IoT solutions for practical engineering applications.
Strength
Designing end-to-end intelligent applications to solve societal and industrial problems. Focus areas include computer vision-driven accessibility technologies, such as real-time sign language detection, and translating models into interactive digital products.
Selected Work
A Flask-based web application that translates Bangla sign language characters and digits into written Bangla in real-time to promote inclusive communication.
Selected Research
Our findings show that using residual blocks is a very effective approach for detecting and classifying white blood cells (WBCs). As residual block offers lossless data restoration, We built a custom CNN model with residual blocks and achieved an accuracy of around 97.65%. We also tested our dataset (Raabin-WBC [4]) on other models that use depthwise separable convolutions, like MobileNet, InceptionNet, and XceptionNet. After fine-tuning, these models achieved accuracies of 97.85%, 98.40%, and 98%, respectively.
This paper introduces a real-time system for detecting and translating Bangla sign characters into written Bangla using an optimized deep learning model deployed within a Flask-based web interface.
This work introduces CyberBiLSTM, a deep learning model trained on a newly aggregated and diverse dataset of over 244,112 SQL injection queries. The model achieved a test accuracy of 98.43% and a precision of 99.79%, outperforming five benchmark classifiers. In particular, these results were obtained in just 15 epochs, demonstrating rapid convergence and strong generalization. The dataset’s diversity—sourced from seven public repositories and one proprietary corpus—enabled the model to learn complex adversarial patterns beyond simple token-level features. Comparative evaluations and ROC/confusion matrix analyses confirm CyberBiLSTM’s superiority in both precision and robustness. This foundation sets the stage for real-world deployment and future extensions, including explainability, zero-day attack detection, and integration into developer-facing security tools.
Experience
2025
Ukil Chamber
Co-founded and established Ukil Chamber to provide accessible professional services. This startup won the University Innovation Hub Pre-Seed Funding.
2024
Web Premium Solution
Leading development and business strategy for a full-time premium web solutions agency.
2025
Honda Bangladesh
Managed student engagement and promotional activities across university channels.
Skills Snapshot
Deep learning, medical imaging, model experimentation, and vision-driven interfaces.
Signal-aware engineering, sensor-connected prototyping, and hardware-software integration.
Product-facing software systems, APIs, and interfaces that stay grounded in real use.
Everyday tools that support shipping, experimentation, and research workflows.
Recognition
Programming Contest Organised by Department of ETE
2022Champion at Programming Contest 2022, Organised by the Department of ETE, RUET for 20 series.
University Innovation Hub Program (UIHP), Cohort 4 — University Innovation Programme, powered by DEIED Project, BHTPA
2025Secured 3rd place in UIHP Cohort 4 with Team Ukil Chamber and received a BDT 40,000 innovation grant through the University Innovation Programme, powered by the DEIED Project, BHTPA.
University Innovation Programme, powered by DEIED Project, BHTPA
2026Received the Best Pitch / Best Presentation Award as a member of Team Ukil Chamber, recognized for delivering a clear, convincing, and well-structured presentation at the University Innovation Programme.
2025 1st International Conference on QPAIN , IEEE Photonics Society Bangladesh Chapter
2025Presented the research paper “Real-Time Detection and Translation of Bangla Sign Language Characters Using Deep Learning” at QPAIN 2025 and received a Certificate and Token of Appreciation in recognition of the contribution.
5th International Conference on ICECTE 2026, RUET
2026Presented the research paper “CyberBiLSTM: A Bidirectional LSTM Architecture for Cybersecurity via High-Precision SQL Injection Attack Detection” at the 5th International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE 2026).
5th International Conference on ICECTE 2026, RUET
2026Presented the research paper “CNN-Transformer Fusion Network for Multi-Class Leaf Disease Classification” at the 5th International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE 2026).
27th International Conference on Computer and Information Technology (ICCIT 2024), IEEE
2024Presented the research paper “Residual Block-Driven CNN for Accurate White Blood Cell Image Analysis and Classification” at the 27th International Conference on Computer and Information Technology (ICCIT 2024).
27th International Conference on Computer and Information Technology (ICCIT 2024), IEEE
2026Presented the research paper “A Hybrid Approach for Accurate Brain Tumor Detection Using Deep Learning Techniques” at the 27th International Conference on Computer and Information Technology (ICCIT 2024).
27th International Conference on Computer and Information Technology (ICCIT 2024), IEEE
2026Presented the research paper “Enhancing Software Defect Prediction Accuracy Using Stacking Ensemble Model” at the 27th International Conference on Computer and Information Technology (ICCIT 2024).
Department of Electronics & Telecommunication Engineering, RUET
2026Received of Appreciation