Najibul Haque Sarker
About me
If you are in a hurry
- Hi I am Najib - Upcoming CS graduate student @ Virginia Tech, MLE @ IQVIA, Research Intern @ Xulab, Fresh Graudate @ BUET CSE.
- Passionate about Deep Learning Research and Competitions. Specifically in the fields of Computer Vision, Video-Language and Multi-Modality.
- Author of 4 papers
- Participated and won in multiple deep learning competitions. Received rank of Kaggle Competitions Master in Kaggle.
- CGPA 3.95/4. Receievd the Dean's List scholarship for academic excellence.
If you have some time
I am Najib (he/him), a fresh graduate from the Department of CSE, BUET. I am currently working as a Machine Learning Engineer in IQVIA which is a human data science company. I have a deep interest for Deep Learning research and competitions. I am the author of 4 published papers. I have also achieved the rank of Competitions Master in Kaggle. My research interests lie in the fields of Computer Vision, Vision-Language as well as Multi-Modality. My current career goal is to pursue a graduate program in my field of interest.
I am working as a research intern for XuLab at Carnegie Mellon University, working under the supervision of Dr Min Xu and guidance of postdoctorial researcher Ali Dabouei on joint Vision-Language tasks. I have worked on two different projects till now. From the first project, I have a first author paper published titled Detecting anomalies from liquid transfer videos in automated laboratory setting which is a work on Video Anomaly Detection. From the second project, I have a co-first author paper titled Leveraging Generative Language Models for Weakly Supervised Sentence Component Analysis in Video-Language Joint Learning which is on Video-Language tasks and has been accepted at CVPRW 2024.
My undergraduate thesis project was under Dr. M. Sohel Rahman through which I developed a scheme utilizing forward diffusion guided reconstruction task. Based on the results of the thesis, we have published a paper titled Forward Diffusion Guided Reconstruction as a Multi-Modal Multi-Task Learning Scheme which was accepted for oral presentation in ICIP 2023. Furthermore, I worked under the supervision of Dr. Shaikh Anowarul Fattah in the task of Synthetic Image Detection and based on our results we published a paper titled ArtiFact: A Large-Scale Dataset with Artificial and Factual Images for Generalizable and Robust Synthetic Image Detection in ICIP 2023.
I am currently working as a Machine Learning Engineer at IQVIA. I am part of the Next Best Action Machine Learning Team, where my day to day work ranges from PoC (Proof of Concept) work of our upcoming products and maintaining our existing algorithm codebase.
Work Experience
-
Machine Learning Engineer
Next Best Action ML Team
IQVIA
June 2023 — Present -
Research Internship
Computational Biology Department
Carnegie Mellon University
Jan 2022 — PresentWorking under the supervision of Dr Min Xu on Computer Vision and Vision-Language projects. Previously worked on video anomaly detection, currently working on video grounding and moment retrieval.
Publications
-
Leveraging Generative Language Models for Weakly Supervised Sentence Component Analysis in Video-Language Joint Learning
Zaber Ibn Abdul Hakim, Najibul Haque Sarker, Rahul Pratap Singh, Bishmoy Paul, Ali Dabouei, Min Xu Accepted at CVPRW 2024 Same contribution as 1st author Computer Vision, Vision-Language, Multi ModalitySecond research intern project under the supervision of Dr Min Xu and project leader Ali Dabouei. In this work, we try to enhance video-language joint learning tasks by incorporating comprehension about significance of sentence components in the context of video-text analysis. Specifically, we utilize LLMs to generate component targeted negative samples which we use for contrastive learning along with an additional adaptive negative importance estimation module. This paper was accepted at CVPRW 2024 (7th Multimodal Learning and Applications Workshop).
| Article | -
Forward Diffusion Guided Reconstruction as a Multi-Modal Multi-Task Learning Scheme
NH Sarker, MS Rahman 2023 IEEE International Conference on Image Processing (ICIP), 3180-3184 1st Author Publication Computer Vision, Medical Imaging, DiffusionThis is based on my undergraduate thesis. Worked under the supervision of Dr. M. Sohel Rahman to develop a novel multi-task mechanism utilizing the forward diffusion process for segmenting brain MRI images. The work was accepted for oral presentation in ICIP 2023.
| Article | Presentation | Preprint | -
ArtiFact: A Large-Scale Dataset with Artificial and Factual Images for Generalizable and Robust Synthetic Image Detection
M. A. Rahman, B. Paul, N. H. Sarker, Z. I. A. Hakim and S. A. Fattah 2023 IEEE International Conference on Image Processing (ICIP), 2200-2204 Same contribution as 1st author Image Generation, Synthetic Image DetectionThis paper is based on our results of IEEE VIP CUP 2022: Synthetic Image Detection Challenge where my team ranked 1st in LB. In this work, to assess the generalizability and robustness of synthetic image detectors in the face of real-world impairments, we presents a large-scale dataset1 named ArtiFact, comprising diverse generators, object categories, and real-world challenges. We propose a multi-class classification scheme combined with a filter stride reduction strategy that addresses social platform impairments and effectively detects synthetic images from both seen and unseen generators. This work was done under the supervision of Dr. Shaikh Anowarul Fattah and the paper was accepted for poster presentation in ICIP 2023.
| Article | Github | -
Detecting anomalies from liquid transfer videos in automated laboratory setting
NH Sarker, ZA Hakim, A Dabouei, MR Uddin, Z Freyberg, A MacWilliams, J Kangas, M Xu Frontiers in Molecular Biosciences 10, 1147514 1st Author Publication Video Anomaly Detection, Object TrackingFirst research intern project under the supervision of Dr Min Xu and project leaders Mostafa Rafid Uddin and Ali Dabouei In this work, we address the problem of detecting anomalies in a certain laboratory automation setting through utilizing practical human-engineered feature extraction method to detect anomalies from liquid transfer video images. The paper was accepted in the journal Frontiers in Molecular Biosciences.
| Article | -
Syn-Att: Synthetic Speech Attribution via Semi-Supervised Unknown Multi-Class Ensemble of CNNs
M. A. Rahman, B. Paul, N. H. Sarker, Z. I. A. Hakim and S. A. Fattah Same contribution as 1st author Signal Processing, Synthetic Speech AttributionThis paper is based on our results of IEEE Signal Processing Cup 2022: Synthetic Speech Attribution Challenge where we became the Winners. The challenge was to detect synthetic speech from natural ones and also identify the algorithm behind the fake speech. In this work, a detector network is proposed that transforms the audio into log-mel spectrogram, extracts features using CNN, and classifies it between five known and unknown algorithms, utilizing semi-supervision and ensemble to improve its robustness and generalizability significantly. This work was done under the supervision of Dr. Shaikh Anowarul Fattah.
| Article | Github |
Education
-
B.Sc. in Computer Science and Engineering
Bangladesh University of Engineering and Technology
April 2018 - May 2023CGPA: 3.95/4.00
-
Higher Secondary School Certificate (HSC)
Notre Dame College
2017GPA: 5.00/5.00
-
Secondary School Certificate (SSC)
St Joseph Higher Secondary School
2015GPA: 5.00/5.00
Technical Skills
-
Programming Languages
Python, C++, JavaScript, Java, C#, SQL, Bash, CSS, Latex
-
Frameworks
Pytorch, Tensorflow, Keras, Sklearn, React, Bootstrap, Django
-
Tools
AWS Sagemaker, Git, Wandb, Trello, Oracle, PostgreSQL