If we knew what it was we were doing, it would not be called research, would it?
Brief Bio
I am currently a PhD candidate at Department of Computer Science in Stony Brook Univeristy. From fall 2018 I’m working in the Computer Vision Lab under the supervision of Prof. Dimitris Samaras.
In Summer and Fall 2025, I interned as an Applied Scientist in Amazon. Where I applied reinforcement learning and Chain-of-Thought to train a multimodal agent (MLLM) to use a set of tools (foundational models) to find visual defects in customer uploaded images. We improved the visual defect classification performance by 6.1%.
Prior to this, I received a Research grant from Zebra Technologies (2022 to 2024), where I worked on deploying large Transformer based models in memory and computation constrained grocery self check out centers. The focus was on object detection from partial object views. This deployment reduced the number of fake transactions by 8.9 percent.
Prior to this, I received my master degree from National Institute of Technology, Calicut, India in 2016 advised by Prof. Sudeep K S.
Research Interest: My research interests span agent AI with reinforcement learning and multimodal reasoning, vision language models for segmentation using images and text, semi-supervised segmentation with limited labeled data, and applied machine learning and computer vision for real world systems.
Industry Experience: I have worked as Senior Software Developer in Cisco for 2 years and prior to that I have almost two years experience in Wipro Technologies as a Project Engineer.
Resume(2025.11)
Contact
- Email : phowlader[at]cs.stonybrook.edu
- Room 138, New Computer Science Building, Stony Brook University, Stony Brook, NY 11794-2424
Publications
- Prantik Howlader, Hoang Nguyen-Canh, Srijan Das, Jingyi Xu, Hieu Le and Dimitris Samaras. “CORA: Consistency-Guided Semi-Supervised Framework for Reasoning Segmentation”. IEEE Winter Conference on Applications of Computer Vision (WACV 2026) (Paper)
- Prantik Howlader, Srijan Das, Hieu Le and Dimitris Samaras. “Beyond Pixels: Semi-Supervised Semantic Segmentation with a Multi-scale Patch-based Multi-Label Classifier”. In European Conference on Computer Vision (ECCV 2024) (Paper)
- Prantik Howlader, Hieu Le and Dimitris Samaras. “Weighting Pseudo-Labels via High-Activation Feature Index Similarity and Object Detection for Semi-Supervised Segmentation”. In European Conference on Computer Vision (ECCV 2024) (Paper)
- Prantik Howlader, Vu Nguyen, Le Hou, Rajarshi Gupta, Dimitris Samaras and Joel Saltz. “Few Shot Hematopoietic Cell Classification”. In Medical Imaging with Deep Learning (MIDL 2023) (Paper)
- Ayush Kumar, Prantik Howlader, Rafael Garcia, Daniel Weiskopf and Klaus Mueller. “Challenges in Interpretability of Neural Networks for Eye Movement Data”. In ACM Symposium on Eye Tracking Research and Applications (ETRA 2020) (Paper)
- Prantik Howlader1, Aditya Chattopadhyay1, Anirban Sarkar1 and Vineeth N. Balasubramanian.“Grad-CAM++: Generalized Gradient-based Visual Explanations for Deep Convolutional Networks,” IEEE Winter Conference on Applications of Computer Vision (WACV 2018) (Paper)
(1 All authors have contributed equally) - Prantik Howlader, Kuntal Kumar Pal, Alfredo Cuzzocrea, and S. D. Madhu Kumar. “Predicting Facebook-Users’ Personality based on Status and Linguistic Features via Flexible Regression Analysis Techniques,” In Proceedings of ACM SAC Conference (SAC 2018) (Paper)
- Prantik Howlader and K.S. Sudeep.“Degree Centrality, Eigen vector Centrality and relation between them in Twitter,” IEEE International Conference on Recent Trends in Electronics, Information & Communication Technology (RTEICT 2016) (Paper)