Baris Sarper Tezcan

I am a Ph.D. student in Computer Science at Purdue University, advised by Prof. Daniel Aliaga. My research interests span computer vision, vision-language learning, generative AI, and remote sensing. I am particularly interested in visual foundation models for geospatial and multimodal understanding, including open-vocabulary recognition, segmentation, visual reasoning, synthetic data generation, and evaluation beyond fixed label sets.

Before joining Purdue, I received my B.Sc. in Computer Engineering from Middle East Technical University (METU), where I also pursued a double major in Mathematics. During my undergraduate studies, I worked on weakly supervised semantic segmentation with Prof. R. Gokberk Cinbis and developed a broader interest in data-efficient learning, representation learning, and generative models.

Profile photo

Publications

Conflated Inverse Modeling to Generate Diverse and Temperature-Change Inducing Urban Vegetation Patterns
Baris Sarper Tezcan, Hrishikesh Viswanath, Rubab Saher, and Daniel Aliaga
CVPR 2026 EarthVision Workshop
Can Your Model Separate Yolks with a Water Bottle? Benchmarking Physical Commonsense Understanding in Video Generation Models
Enes Sanli*, Baris Sarper Tezcan*, Erkut Erdem, and Aykut Erdem (*equal contribution)
MILA World Modeling Workshop, 2025
Extended version under review via ACL Rolling Review, 2026
Advancing Multimodal LLMs by Large-Scale 3D Visual Instruction Dataset Generation
Liu He, Yizhi Song, Ruiqi Xu*, Baris Sarper Tezcan*, Xiao Zeng, Albert Y. C. Chen, Lu Xia, Shashwat Verma, Sankalp Dayal, Min Sun, Cheng-Hao Kuo, and Daniel Aliaga (*equal contribution)
Under review at Machine Vision and Applications, 2026

Projects

Weakly Supervised Semantic Segmentation Research
Baris Sarper Tezcan, supervised by R. Gokberk Cinbis
Middle East Technical University, 2022-2023
report

Analyzed confident region thresholds in AffinityNet-style WSSS pipelines on PASCAL VOC 2012. The project focused on selecting foreground and background pseudo-label thresholds more systematically under image-level supervision.

DiffDis ICCV 2023 Paper Implementation
Baris Sarper Tezcan
Deep Generative Models Course, 2024
code / paper

Implemented "DiffDis: Empowering Generative Diffusion Model with Cross-Modal Discrimination Capability" as a working reproduction for a course project, filling in missing implementation details from the paper.

RoboDetection
Baris Sarper Tezcan and team
METU Computer Engineering Graduation Project, 2024
project page

Led a five-person team building a robot dog simulation for disaster response, including remote control, ROS networking, autonomous person detection and tracking, and real-time 2D map generation.

Experience

2023 Computer Vision / Machine Learning Intern, Kuartis. Studied weather classification, semantic occupancy prediction, horizon detection, OpenCV trackers, dataset curation, NAS, and hyperparameter optimization. report
2022 Research Intern, ROMER Research Center. Implemented and compared feature-based image matching methods including SIFT, SURF, FAST, ORB, and SuperGlue, with RANSAC-based outlier removal. report

Awards

2024 METU Graduation Project Competition, 3rd Place
2021 TEKNOFEST Artificial Intelligence in Transportation Competition, 2nd Place
2020 TUBITAK Unmanned Aerial Vehicle Competition, 7th Place