Stony Brook University Computer Science

The Conference on Computer Vision and Pattern Recognition (CVPR) 2026, regarded as one of the most prestigious and highest-impact conferences, has recognized Prof. François Rameau as an Outstanding Reviewer.  This recognition was given to the top 5% of reviewers for their high-quality reviews among 17,491 reviewers.  This is the third time Prof. François Rameau has received the Outstanding Reviewer  Award from CVPR. Link: https://x.com/CVPR/status/2056432742208876945?s=20

An undergrad research project that began in the SUNY Korea CS Department has led to a paper accepted at the International Conference on Pattern Recognition, ICPR, a major international venue in pattern recognition and computer vision. Yerin Cheon started the work as part of her CSE487 research project, co-supervised by Prof. François Rameau and Prof. Aruna Balasubramanian from Stony Brook University, who was visiting SUNY Korea at the time. After graduating from SUNY Korea, Yerin joined Stony Brook University for her master’s degree, where she continued developing the project in Prof. Aruna Balasubramanian’s group. The resulting paper, “Dual-Foundation Models for Unsupervised Domain Adaptation,” is the outcome of this continued collaboration. The paper addresses a core challenge in robotics and autonomous vehicles: safely navigating an environment requires understanding the surrounding scene. This is often achieved through semantic segmentation, where each pixel in an image is assigned a category label. Training such deep learning models typically requires large amounts of pixel-level annotation, and the resulting models often do not transfer well to new environments. Unsupervised domain adaptation aims to mitigate that issue, but existing methods still struggle to produce reliable pseudo-labels and robustly align features across domains. Yerin’s approach tackles this limitation by leveraging vision foundation models to improve both pseudo-label refinement and feature alignment. The work is a strong example of how undergraduate research at SUNY Korea can develop into peer-reviewed international publications and continue through collaboration with Stony Brook University.   Research Information · Paper Title: Dual-Foundation Models for Unsupervised Domain Adaptation · Conference: ICPR 2026 · Authors: Yerin Cheon, Aruna Balasubramanian, and Francois Rameau · Project page: https://github.com/ycheon1101/DFUDA · Paper: https://arxiv.org/pdf/2605.03365

  Prof. François Rameau's research team has revealed a major privacy vulnerability in AI-based visual localization models. Their research paper, titled “Seeing Through the Weights: Privacy Leakage in Scene Coordinate Regression,” has been accepted to ECCV 2026, one of the world's most prestigious venues in computer vision and AI. The conference will be held in Malmö, Sweden, from September 8 to 12, 2026. Certain neural network-based visual localization models have long been considered naturally privacy-preserving, since they encode a scene implicitly in their parameters rather than storing images or 3D maps. This work shows that this assumption does not hold: trained models can still leak sensitive information about the environments used to train them.   This research was conducted by Ph.D. student Oleksii Nasypanyi and M.S. student Jaemin Cho, under the supervision of Prof. Utku Ozbulak, Prof. Byungkon Kang, and Prof. François Rameau. The project also reflects a close collaboration within the Incheon Global Campus, bringing together researchers from SUNY Korea, Ghent University Global Campus, and George Mason University Korea.   The team shows that an attacker does not need any images of the target scene to extract information from the model. By simply feeding it random, unrelated images and collecting its outputs, they can gradually reconstruct both the structure and the appearance of the environment the model was trained on. The attack works under different levels of access to the model, from full access to more restricted settings where only its outputs are visible By identifying this vulnerability, the research highlights the need for stronger privacy-aware design and evaluation in AI-based spatial intelligence systems. Research Information Paper Title: Seeing Through the Weights: Privacy Leakage in Scene Coordinate Regression Conference: ECCV 2026 Authors: Oleksii Nasypanyi, Jaemin Cho, Utku Ozbulak, Byungkon Kang, François Rameau Project page: https://jaeminch0.github.io/seeing-through-the-weights-privacy-leakage-in-scene-coordinate-regression/ Paper: https://arxiv.org/abs/2606.31164

  A research paper led by Ph.D. student Si Yong Lee has been accepted for publication at the 2026 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), the world's leading international conference in medical AI and computer-assisted healthcare technology, to be held in Strasbourg, France, from September 27 to October 1, 2026. MICCAI is recognized as a top-tier venue with a highly competitive review process, and acceptance represents a significant research achievement. The work was co-authored with graduate students Ryangjin Lee, Hawon Park, and Yoora Kim, and was conducted under the supervision of Prof. Yoon Seok Yang and Prof. Byungkon Kang, reflecting the CS Department's strong culture of close faculty mentorship and student-driven research.   The paper, "SpikeMamba: Spike-Driven State Space Models for Energy-Efficient Biomedical Sequence Modeling," presents the first architecture to bring Mamba state space models entirely into the domain of spiking neural networks, enabling brain-inspired, event-driven computation for biomedical signals such as ECG and EEG. SpikeMamba matched or exceeded the accuracy of state-of-the-art models while consuming over 30 times less energy, opening a practical path toward always-on health monitoring on battery-powered wearable devices and neuromorphic processors such as MindCore.   This achievement highlights the CS Department's growing strength in neuromorphic computing and energy-efficient AI, and its commitment to research with real-world clinical impact.

The Department of Computer Science at SUNY Korea successfully hosted the Spring 2026 Undergraduate Research Opportunities Program (UROP) on May 27, 2026, in the SUNY Korea Student Lounge. The event brought together approximately 70 students, faculty, and staff to celebrate undergraduate research and provide students with an opportunity to present the projects they have been developing throughout the semester.   The program featured both oral presentations and poster sessions covering a wide range of computer science topics. Undergraduate researchers shared their work with faculty members and fellow students, explaining their research ideas, methodologies, and results. The event encouraged meaningful academic discussions while allowing participants to explore the diverse research activities taking place within the department.   Beyond showcasing research outcomes, the UROP served as a valuable learning experience for student presenters. By communicating their work to a broader audience and receiving feedback from faculty and peers, students strengthened both their technical presentation skills and their understanding of research. The event also allowed attendees to discover ongoing research projects and inspired many students to consider participating in research in the future.   The Spring 2026 UROP received overwhelmingly positive feedback from participants, earning an overall satisfaction rating of 4.59 out of 5.0. Attendees particularly appreciated the variety of student presentations and the opportunity to learn about research being conducted across the department. Building on the success of this semester's event, the Department of Computer Science will continue to support undergraduate research initiatives that foster innovation, collaboration, and academic excellence.

Three undergraduate students (Dongyeong Kim, Gilho Kim, and Myeongkjun Kim), together with one Ph.D. student (Si Yong Lee), have published a research paper in IEEE Access, a Q1-ranked international journal, demonstrating an outstanding educational and research achievement. This work was conducted under the supervision of Prof. Yoon Seok Yang and reflects the strong research-oriented educational environment fostered through close faculty mentorship. The research focuses on improving the training and inference efficiency of spiking neural network (SNN)–based object detection models for low-power, real-time edge AI applications. To address the high computational cost of conventional CNN-based detectors such as YOLO, the group enhanced a Sigma–Delta spiking YOLOv3 framework by incorporating advanced training and optimization techniques. The proposed methods improved convergence, detection accuracy, and energy efficiency, and were validated using multiple real-world datasets. In addition, the trained model was successfully deployed on Intel’s Loihi 2 neuromorphic platform, demonstrating its feasibility for real-time and low-power operation. Through this project, undergraduate students gained hands-on experience in advanced AI research, from algorithm development to deployment on a neuromorphic hardware platform. This work demonstrates the value of active undergraduate participation in research and the CS Department’s commitment to integrating education and cutting-edge research.

  Two CS undergraduates, Juyoung Yun and Sol Choi, were the lead authors on a paper that won the Best Paper Award at the 2025 International Conference on Neural Information Processing (ICONIP), a long-running and well-regarded conference in the field of AI. The other authors were their supervisors: Profs. Francois Rameau, Prof. Byungkon Kang, and Prof. Zhoulai Fu. The paper is “Revisiting 16-bit Neural Network Training: A Practical Approach for Resource-Limited Learning”. Modern AI models are usually trained in 32-bit “full precision,” which is accurate but expensive in terms of memory and computation. Their work shows that one can safely train neural networks entirely in 16-bit precision, without the usual 32-bit “safety net,” even on modest hardware. A practical recipe, validated by theory and extensive experiments, is given for choosing the right settings so that training remains stable and the final model quality matches (and sometimes even rivals) 32-bit training. This is particularly important for settings where computing resources are limited, such as smaller labs, edge devices, or student projects that cannot rely on large GPU clusters. By cutting the precision in half while preserving performance, our approach makes state-of-the-art neural network training more accessible and more energy-efficient. This efficiency and practicality were key reasons the paper received the award.

Brain-inspired Computing Lab (BCL), led by Prof. Yoon Seok Yang, was selected to participate in the 2025 Preliminary Startup Package program organized by the Ministry of SMEs and Startups of Korea. BCL’s selected project, “Next-Generation AI Neuromorphic Processor Semiconductor Design”, is about developing a new type of on-device AI chip for edge computing that works more like the human brain. Most of today’s AI systems use GPUs, which are powerful but consume a lot of electricity and cost a great deal to run. As AI models get bigger and more complex, this problem is becoming even more serious. BCL aims to create a processor based on neuromorphic computing, which mimics the way neurons and synapses in the brain work. This makes it possible to run AI much faster and with far less energy, which is especially important for edge devices like self-driving cars, robots, and smart devices that need to work in real time.  In the project, BCL plans to design the chip itself, build a test version using an FPGA board, and make sure it can handle well-known datasets like ImageNet. BCL are also creating easy-to-use software tools so that developers can work with their chips just like they would with existing platforms such as TensorFlow or PyTorch. In the future, the goal is to make these chips available for many different applications and contribute to AI technology that is not only more powerful but also more energy-efficient and accessible.

  The department is pleased to welcome two new faculty members whose teaching and professional experiences further enrich our department's commitment to high-quality education and innovation. Prof. JiHyun Jeong joins the department as an Assistant Professor. She recently completed her Ph.D. in Information Science from Cornell University, where her research focused on Human-AI interaction, trust, and collaboration. Her dissertation explored how autonomous agents can balance performance with social considerations when interacting with humans. Prof. Jeong brings a strong interdisciplinary background in human-computer interaction, AI-driven social behaviors, and team cooperation technologies. She has also worked as a PhD intern at Motional and has received multiple awards, including an Honorable Mention at ACM DIS and a DARPA grant as co-author. Her research expands the department's strength in socially aware AI, human-agent interaction, and computational models of cooperation. Prof. James Finn joins as a Teaching Professor, bringing a wealth of international teaching and industry experience. He holds a Ph.D. in Computer Science from Princeton University and has taught computer science and engineering at institutions such as Stanford University, San Francisco State University, and King Mongkut’s University of Technology Thonburi (Thailand). His background includes roles as a software engineer at Yelp, technical trainer for Hewlett Packard and Tibco Software, and senior technical writer at MacUser Magazine. Prof. Finn’s extensive experience in technical instruction and global education will greatly enhance our students’ learning experience, particularly in programming and foundational CS courses.

    Prof. François Rameau and his team from SUNY Korea participated in the 5th Annual Nepal AI School (ANAIS), held from December 27, 2024, to January 6, 2025, in Kathmandu, Nepal. On January 1st, Prof. Rameau delivered a captivating lecture on the transformative field of Computer Vision, offering participants an in-depth journey through pivotal research papers and groundbreaking advancements that have shaped the field, particularly in 3D reconstruction. His session provided invaluable insights into the historical and technical context of Computer Vision, emphasizing how cutting-edge developments are empowering machines to perceive, interpret, and interact with the world in unprecedented ways. Participants praised the lecture for its clarity and relevance, which bridged foundational concepts with real-world applications. The Annual Nepal AI School, organized by the Nepal Applied Mathematics and Informatics Institute for Research (NAAMII), is a premier platform in South Asia, bringing together international scientists, faculty, and participants for 11 intensive days of AI-focused lectures, labs, and project work. The program covers a comprehensive range of topics, from fundamental principles to the latest advancements in AI research and applications, fostering collaboration and innovation in the field. https://nepalschool.naamii.org.np/home https://www.linkedin.com/posts/naamiinepal_anais2024-naamii-researcheducateinnovate-activity-7284120674200178688-rb6P?utm_source=social_share_send&utm_medium=member_desktop_web  

  Jaemin Cho, a graduate student from the Computer Science MS program, achieved a remarkable feat by winning the prestigious Hackathon Challenge for Layering Detection in Anti-Money Laundering. This challenge, hosted during the Annual Nepal AI School (ANAIS) 2024 in Nepal on December 26th,2024 to January 6th, 2025, called for innovative solutions to detect layering in financial transactions—a critical step in combating money laundering. The competition focused on creating advanced methodologies to identify and prevent layering, a key stage in the money laundering process. The significance of this challenge extends beyond the technical aspects. It aims to prevent financial crimes by identifying illicit transactions early, ensure compliance with global regulations and laws, and protect economies by mitigating illicit activities that can destabilize financial systems. As the winner, Jaemin has earned accolades and the opportunity to be hired by F1Soft International, one of Nepal’s leading financial technology companies. This recognition highlights the practical and transformative applications of his skills in financial technology. The ANAIS 2024 event was a vibrant platform, featuring 191 participants and 22 speakers representing 18 countries, fostering a truly global exchange of ideas. Twenty-two competing teams, each comprising five members, showcased their innovative solutions to challenges, including the F1Soft Money Laundering Challenge and the Startup Challenge. Participants had the opportunity to present their work, sharpen their skills, and contribute to impactful problem-solving on a global scale. The event’s theme, "Big Things Can Happen from Small Places," underscored the potential for groundbreaking ideas to emerge from any part of the world. ANAIS 2024 was made possible through the support of its sponsors and partners. The title sponsor, F1Soft International Pvt. Ltd., and the silver sponsor, YoungInnovations Pvt. Ltd., provided vital backing for the event. The venue partner, Premier International IB Continuum School, hosted the event, while the AI Hackathon received additional sponsorship from UK International Development, The Asia Foundation, Data for Development in Nepal, eSewa Ltd., and Fonepay. Further support from Google, SecurityPal, and Fusemachines reinforced the collaborative spirit of this initiative. Jaemin’s accomplishment not only showcases his talent and dedication but also brings pride to our academic community. His success serves as an inspiration to all students, proving that with hard work, innovation, and determination, transformative contributions can be achieved on a global stage.   https://nepalschool.naamii.org.np/home https://www.linkedin.com/posts/naamiinepal_anais2024-naamii-researcheducateinnovate-activity-7284739753776660480-MO5w?utm_source=social_share_send&utm_medium=member_desktop_web https://www.linkedin.com/posts/naamiinepal_anais2024-naamii-researcheducateinnovate-activity-7285464529910603776-VdZT?utm_source=social_share_send&utm_medium=member_desktop_web

Prof. Kang’s latest publications include a paper in the proceedings of the 2024 International Conference on Information and Knowledge Management (CIKM). This is a premier AI conference well-known for its papers on data mining, machine learning, and information management. It is recognized by many academic institutions, including NRF in Korea, as a class-A CS conference (CORE rank A).  The paper is about a machine learning algorithm that bypasses a shortcoming of AutoDiff, a core component in deep learning that enables us to train complex models. When training a model that uses linear algebraic routines, such as determinants, AutoDiff will give suboptimal performance. The paper proposes a reparameterization algorithm that yields the correct learning scheme as well as potentially saving more time by taking advantage of the parallelizability of the new formulation. A second paper, with Ph.D student Joonkyu Han as the main contributor, was published in the proceedings of the 29 th IEEE Pacific Rim International Symposium On Dependable Computing (PRDC). This is a well-regarded conference that showcases a variety of novel papers on software reliability, safety, security, and dependability. The paper deals with the use of visual and spatial information to authenticate a user during a login process. A fall-back authentication is an extra step of authentication taken when the user forgets his/her password. Typically-used secure question or email verification approaches have their shortcomings in terms of memorability and security. This work provides a way to authenticate a user by exploiting visual navigation through a virtual map; that is, a successful navigation through a map constitutes a successful login. Humans are particularly well-suited to remembering visual navigation, a fact that motivates and justifies our approach. Prof. Kang was also recognized as an Outstanding Reviewer for the 35th British Machine Vision Conference (November 2024), a highly ranked conference on ahigh-ranking computer vision.

The Conference on Computer Vision and Pattern Recognition (CVPR) 2024, regarded as one of the most prestigious and highest-impact conferences, has recognized Prof. Rameau as an Outstanding Reviewer.  This recognition was given to the top 2% of reviewers for their high-quality reviews (as judged by their Area Chairs) among 9,872 reviewers.  This is the second time Prof. Rameau has received the Outstanding Reviewer Award from CVPR. Link: https://x.com/CVPR/status/1793616950314369239

  Professors Niranjan Balasubramanian and Michalis Polychronakis, from the Department of Computer Science, have each received Amazon Research Awards to further advance their fields of research.  Balasubramanian’s research focuses on the potential of Large Language Models (LLMs) for autonomous execution of complex tasks. He will use the Amazon funding to create a controlled environment, a complex task testbed, where LLMs can be rigorously evaluated. This testbed features innovative assessment criteria beyond typical accuracy metrics, a sandbox execution environment with mock APIs, and natural language descriptions of complex goals.   This research bridges the gap between theoretical promise and real-world implementation. By developing a controlled environment for LLMs, Balasubramanian aims to unlock their potential while ensuring safety. His work contributes to advancing AI technologies and addressing real-world challenges in a thoughtful and systematic manner.  Polychronakis’ Amazon funding will allow him to continue to explore ways to improve software security and enhance memory safety. His research aims to address the challenges posed by memory corruption vulnerabilities, which are still a major source of system compromise and malware infection. Despite the advantages of modern memory-safe languages like Go and Rust, most existing software are still written in memory-unsafe languages like C and C++. The familiarity of developers with C and C++, vast code bases in these languages, and their efficiency hinder efforts to migrate to memory-safe alternatives.  To address this issue, Polychronakis is developing SafeTrans, a system that automates the conversion of existing C/C++ code to Rust. Rust with its memory safety features and low runtime overhead, is a great candidate to replace memory-unsafe languages in critical systems. SafeTrans seeks to accelerate the adoption of memory-safe languages by automating elements of the migration process while lowering the risk of memory-related vulnerabilities. His research helps the larger goal of increasing software security and making systems more resistant to modern vulnerabilities.  Both researchers received approximately $100k in funding which includes Amazon credits.  The Amazon Research awards recognize the innovative contributions of Niranjan Balasubramanian and Michalis Polychronakis, and their research assistants. Both initiatives demonstrate a commitment to innovation and real-world impact, shaping the future of technology.   Written by Sahil Sarna  

  Computer Science students have the opportunity to gain various experiences during their visit to the SBU NY campus.  Four students from the Department of Computer Science at SUNY Korea achieved three meaningful awards in the three biggest Hacks competitions at Stony Brook University in New York. These include “Best Sustainability Hack” in SBU Hopper Hacks 2024, “Best Social Impact” in SBUHacks VI, and “Best Integration of Customer Segmentation Utilizing AI” in Softheon Hack@CEWIT. These events were held online from January 29 to February 4, February 9 to February 11, and March 1 to March 3, 2024, respectively, at Stony Brook University. Biniam Markos, Eunwoo Choi, Ulukbek Aitmatov, and Younwoo Ki from SUNY Korea won the  “Best Sustainability Hack” as one team in SBU Hopper Hacks 2024. Hopper Hacks is a hackathon centered around social good, lasting 24 hours, and open to all students at Stony Brook University. SUNY Korea CS students were also awarded in the same competition in 2022. The same team also won the “Best Social Impact” in SBUHacks VI. SBUHacks extends over 48 hours, encouraging limitless creativity. During SBUHacks, students are tasked with pushing their boundaries and pursuing projects aligned with their interests. Additionally, Biniam Markos, Eunwoo Choi and Younwoo Ki from SUNY Korea, along with Alissa Burich, majoring in Business Management from SBU in New York, won “Best Integration of Customer Segmentation Utilizing AI” in Softheon Hack@CEWIT 2024, which is the second-place prize in this competition. Notably, SUNY Korea students have been consistently recognized in this competition, with another group securing the highest award last year. At Softheon Hack@CEWIT, participating students can develop software and automation solutions tailored for health insurers and governmental health agencies.   Written by Student Reporter, Hyeonha Hwang (hyeonha.hwang@stonybrook.edu)