About Me

Bio

I am a Lecturer in the Computer Science and Engineering Department at BRAC University, Dhaka, Bangladesh. I hold a Bachelor’s degree in Computer Science and Engineering from BRAC University and am pursuing a Master’s degree at the same institution. Passionate about Computer Vision, AI and AI security, I combine expertise in machine learning and full-stack development to drive innovative solutions.

Work Email: ext.irtiza.hossain@bracu.ac.bd
Personal Email: mohammad.irtiza.hossain@gmail.com

Research Interests

• Computer Vision and Machine Learning: Uncertainty and calibration, robustness to distribution shift and tampered data, and efficient models
• 3D Object Detection (LiDAR, BEV representation)
• Video Behavior Recognition & Lightweight Document Automation
• AI Security, Steganography & Privacy

Publications

Predictive Self-Conditioning for Accurate and Calibrated LiDAR 3D Object Detection
IEEE/CVF WACV 2027 (CORE A, Conference Paper - Accepted)
Sami Sadat, Md. Irtiza Hossain, Shreya Adrita Banik, Md Zia Uddin, Md. Golam Rabiul Alam

This paper introduces Predictive Self-Conditioning (PSC), a within-frame feature-level feedback mechanism that converts Pass-1 proposals into confidence-weighted Gaussian heatmaps and injects them as a BEV feature residual before Pass-2, significantly boosting accuracy and calibration for LiDAR 3D object detection.
[Details]

Edge-Native Digitization of Handwritten Marksheets: A Hybrid Heuristic-Deep Learning Framework
QPAIN 2026 (Conference Paper)
Md. Irtiza Hossain, Junaid Ahmed Sifat, Abir Chowdhury, Md. Imran Bin Azad

This paper introduces a resource efficient hybrid framework for digitizing structured handwritten documents, combining a heuristic OpenCV-based pipeline for table detection with a modified lightweight YOLOv8 for handwritten character recognition, achieving real-time performance and high accuracy on edge devices.
[Details] [DOI]

Enhanced CNN Approaches for Multi-Image Embedding in Image Steganography
ICICT 2024 (Conference Paper)
Md. Irtiza Hossain, Samiul Kadir, Farhan Ishraq Fagun, Ishtiaq Samiul, Rafi Zaman Saukhin

This paper proposes novel convolutional neural network (CNN) approaches for embedding multiple images within a single cover image for steganography, enhancing data hiding capacity and security.
[Details] [DOI]