Nazmul Karim
I am actively looking for full-time opportunities in a machine learning researcher/engineering role! Have 5 years of experience in Machine Learning and Signal Processing. I am quick at learning new things and have worked on diverse research projects.
About Me
Hi, I am Nazmul Karim. I completed my Ph.D.in Electrical Engineering and M.Sc. in Computer Engineering from the University of Central Florida (UCF). I was jointly advised by Nazanin Rahnavard (director of LCWNLAB), and Mubarak Shah (director of CRCV). Before that, I got my B.Sc. in Electrical and Electronics Engineering (EEE) from Bangladesh University of Engineering and Technology (BUET) in 2016. I work primarily in computer vision although I have some notable experience in Natural Language Processing (NLP) too. My research interest includes but is not limited to Large Vision-language Models Robustness, Noisy Labels, Domain Adaptation, Continual Learning, Generative AI, Multimodal AI, 3D Scene Generation, etc. I have published at top-tier journals and conferences like IEEE TIFS CVPR MLSP. I also served as a reviewer in CVPR, MLSP, and Optics Journal.
Experience
Graduate Research Assistant
at University of Central Florida (UCF), Orlando, FL, USA
- Timeline
- August 2018 - Present
- Responsibilities
- Developed multiple advanced frameworks for the detection and mitigation of Trojans within deep neural networks. These frameworks are built upon concepts such as the geometry of decision boundary and loss landscape, Fisher Information Matrix (FIM), Neural Masking, etc. Achieved SOTA performance across various settings, e.g. Image Classification, Object Detection, Video Action Recognition, 3D Point Cloud, Natural Language Generation, etc.
- Formulated an innovative training framework for noisy labels at the forefront of the field, resulting in a noteworthy 11% improvement in accuracy. This framework demonstrated comparable efficacy across both synthetic and real-world noisy data scenarios. I have also explored the application of noisy labels in source-free domain adaptation.
- Engineered multiple diffusion-based frameworks with the capacity to facilitate text-to-image, text-to-video, and text-to-3D scene generation. These frameworks can also support zero-shot high-fidelity content generation.
- Devised a framework founded on Generative Adversarial Network (GAN), tailored for the compression of images and videos. Accomplished an impressive compression rate of approximately 100-fold, while impeccably preserving the inherent quality of the visual content.
Applied Scientist Intern
at Amazon Web Services, Seattle, WA
- Timeline
- June 2023 - Sep 2023
- Responsibilities
- Formulating algorithms for Image Matching and Pose Estimation, with a specific emphasis on both indoor and outdoor pose estimation. With the help of foundation models such as CLIP and DinoV2, my goal was to achieve a level of generalization applicable across diverse real-world scenarios
Research Intern
at SRI International, Princeton, NJ
- Timeline
- May 2022 - Aug 2022
- Responsibilities
- Developed a novel source-free domain adaptation technique and achieved on average 2% performance improvement on 7 different tasks covering both image classification and semantic segmentation.
Machine Learning Researcher
at Semion Inc., Dhaka, Bangladesh
- Timeline
- Aug 2016 - April 2017
- Responsibilities
- Development of a tumor detection framework through the utilization of convolutional neural networks. Implementation of text summarization and sentiment analysis methodologies employing Long-Short-Term Memory (LSTM) networks.
Lecturer
at Bangladesh University of Business and Technology (BUBT), Dhaka, Bangladesh
- Timeline
- Aug 2016 - Jul 2018
- Responsibilities
- Served as a mentor for engineering students and an advisor of the University Robotics Club.
Education
Ph.D in Electrical Engineering
from University of Central Florida (UCF)
- Completion Year
- November 2023
- Focus
- Computer Vision, Safe and Robust AI, Representation Learning, Compressive Sensing, and Signal Processing.
- Advisor
- Nazanin Rahnavard, Professor, Department of ECE, UCF
- Co-Advisor
- Mubarak Shah, Director, Center for Research in Computer Vision (CRCV), UCF
M.Sc. in Computer Engineering
from University of Central Florida (UCF)
- Completion Year
- May 2020
B.Sc. in Electrical Engineering
from Bangladesh University of Engineering and Technology (BUET)
- Completion Year
- March 2016
Publications
- Free-Editor: Zero-shot Text-driven 3D Scene Editing
Nazmul Karim(*), Umar Khalid(*), Hasan Iqbal(*), Jing Hua, Chen Chen
European Conference on Computer Vision (ECCV), Italy
- LatentEditor: Text Driven Local Editing of 3D Scenes
Umar Khalid(*), Hasan Iqbal(*), Nazmul Karim(*), Muhammad Tayyab, Jing Hua, Chen Chen
European Conference on Computer Vision (ECCV), Italy
- C-SFDA: A Curriculum Learning Aided Self-Training Framework for Efficient Source Free Domain Adaptation
Nazmul Karim, Niluthpol Chowdhury Mithun, Abhinav Rajvanshi, Han-pang Chiu, Supun Samarasekera, Nazanin Rahnavard
IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2023, CANADA
- UNICON: Combating Label Noise Through Uniform Selection and Contrastive Learning
Nazmul Karim, Mamshad Nayeem Rizve, Nazanin Rahnavard, Ajmal Mian, Mubarak Shah
IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2022, USA
- CNLL: A Semi-Supervised Approach for Continual Noisy Label Learning
Nazmul Karim, Umar Khalid, Ashkan Esmaeili, Nazanin Rahnavard
3rd CLVision Workshop at CVPR 2022, USA
- Odyssey: Creation, Analysis, and Detection of Trojan Models
Marzieh Edraki(*), Nazmul Karim(*), Nazanin Rahnavard, Ajmal Mian, Mubarak Shah
IEEE Transactions on Information Forensics and Security (TIFS) 2021
- RL-NCS: Reinforcement Learning Based Data-driven Approach for Nonuniform Compressed Sensing
Nazmul Karim, Alireza Zaeemzadeh, Nazanin Rahnavard
IEEE Conference on Machine Learning and Signal Processing (MLSP) 2019
- RF Signal Transformation and Classification using Deep Neural Networks
Umar Khalid, Nazmul Karim, Nazanin Rahnavard
SPIE Proceedings of Big Data IV: Learning, Analytics, and Applications 2022
- RODD: A Self-Supervised Approach for Robust Out-of-Distribution Detection
Umar Khalid, Ashkan Esmaeili, Nazmul Karim, Nazanin Rahnavard
The Art of Robustness: Devil and Angel in Adversarial Machine Learning Workshop at CVPR 2022