Computer Vision Researcher | Deep Learning | Human Body Analysis
Website |
Google Scholar |
LinkedIn
I am a computer vision researcher passionate about state-of-the-art deep learning for visual human body analysis. My work focuses on human pose estimation, detection, and segmentation and multi-task learning — pushing the boundaries of in-the-wild human body analysis. I’m especially interested in robust, efficient methods for real-world vision problems and developing open-source solutions for the research community.
- 🔬 Currently working on advanced methods for human pose estimation and multi-body detection.
- 🏆 ICCV 2025 & CVPR 2025 first-author.
- 🌱 I enjoy player tracking in team sports and building methods for dense, complex scenes. I want to make computer vision methods robust enough to use in real-world (eg. sports).
For more, visit my website.
- Languages: Python, (C/C++)
- Core Expertise: Computer Vision, Deep Learning, Multi-task Learning
- Frameworks/Tools: PyTorch, OpenCV, NumPy, MMPose, MMDetection, ViTPose, SAM2
Detection, Pose Estimation and Segmentation for Multiple Bodies: Closing the Virtuous Circle
- Introduces MaskPose: A pose estimator conditioned on segmentation masks for dense scenes.
- Integrates detection, pose estimation, and segmentation in a self-improving loop -- BBoxMaskPose.
- Paper • Website • Hugging Face Models
ProbPose: A Probabilistic Approach to 2D Human Pose Estimation
-
Probabilistic human pose estimation — reliable uncertainty quantification for keypoints.
-
More realistic evaluation method implemented in the ExoCocoTools package.
- Multi-body detection & pose estimation
- Player tracking in team sports (work in progress)
- Robust, reproducible science
When I'm not coding or reading papers, I’m probably:
- Hiking in nature
- Enjoying outdoor adventures
- Exploring new books (high fantasy, psychology, economy, ...)
- Personal Website
- Google Scholar
“Bringing robust vision algorithms from research to real-world impact.”