Arnab Karmakar

arnabk1 [at] uw [dot] edu

Hey there!

I’m Arnab, a Master’s student specializing in Machine Learning and GPU Computing at the Department of Electrical and Computer Engineering, University of Washington, Seattle.

As a Graduate Researcher at the Reasoning, AI, and VisioN (RAIVN) Lab under the guidance of Prof. Ranjay Krishna, my work revolves around Multimodal Machine Learning. My research focuses on efficient representation learning methods for foundation models, improving image and video understanding.

I recently completed my research internship at Brainchip where I developed an efficent embedding compression technique to make low power edge devices faster. During summer, I interned at Truemedia.org where I was responsible for developing end-to-end ML models for detecting deepfake media.

Prior to my Master’s, I was a Senior Research Scientist at the Human Space Flight Center, Indian Space Research Organization (ISRO). In this role, I led the development of a real-time astronaut health monitoring system using Machine Learning. Notably, I led the Crew Training Simulator team and Risk Assessment team for ISRO’s first manned mission, ‘Gaganyaan.’

I am always interested to discuss about the recent developments and future directions in the evolving field of AI. The best projects often begin with a simple conversation.

news

Sep 16, 2024	Joined Brainchip as a Machine Learning Research Intern or Fall 2024!
Jun 10, 2024	Joined Truemedia.org as a Machine Learning Intern or Summer 2024!
Jun 4, 2024	Our GPU Programming project on improving GEMM kernels usig Stream-k won the 2nd best project award from AMD! (ppt) (report)
Mar 16, 2024	Developed a multi-agent personified chatbot mimicking the famous persona of Steve Jobs (link)
Mar 7, 2024	Completed the Project “Analysis and Benchmarking of Prompt Injection Attacks on Back Box Large Language Models (LLMs)” (link)

selected publications

Pose invariant person re-identification using robust pose-transformation gan

Arnab Karmakar, and Deepak Mishra

arXiv preprint arXiv:2105.00930, 2022

Abs Bib HTML PDF

The objective of person re-identification (re-ID) is to retrieve a person’s images from an image gallery, given a single instance of the person of interest. Despite several advancements, learning discriminative identity-sensitive and viewpoint invariant features for robust Person Re-identification is a major challenge owing to large pose variation of humans. This paper proposes a re-ID pipeline which utilizes the image generation capability of Generative Adversarial Networks combined with pose clustering and feature fusion to achieve pose invariant feature learning. The objective is to model a given person under different viewpoints and large pose changes and extract the most discriminative features from all appearances. The pose transformation GAN (pt-GAN) module is trained to generate a person’s image in any given pose. In order to identify the most significant poses for discriminative feature extraction, a Pose Clustering module is proposed. The given instance of the person is modelled in varying poses and these features are effectively combined through the Feature Fusion Network. The final re-ID model consisting of these 3 sub blocks, alleviates the pose dependence in person re-ID. Also, The proposed model is robust to occlusion, scale, rotation and illumination, providing a framework for viewpoint invariant feature learning. The proposed method outperforms the stateof-the-art GAN based models in 4 benchmark datasets. It also surpasses the state-of-the-art models that report higher re-ID accuracy in terms of improvement over baseline.
@article{karmakar2021pose, title = {Pose invariant person re-identification using robust pose-transformation gan}, author = {Karmakar, Arnab and Mishra, Deepak}, journal = {arXiv preprint arXiv:2105.00930}, year = {2022}, }