Mark Weber

Papers

Total Citations

H-Index

About

Mark Weber is a computer vision researcher whose work focuses on efficient, real-time scene understanding. His most notable contribution is the development of a single-shot panoptic segmentation method, which unifies the segmentation of countable objects ("things") and amorphous background regions ("stuff") into a single, non-overlapping output. Unlike prior state-of-the-art approaches that relied on complex, multi-stage pipelines and struggled to reach video frame rates, Weber's end-to-end architecture achieves near real-time performance, processing images at almost video speed. This breakthrough, detailed in his 2020 paper "Single-Shot Panoptic Segmentation," has garnered 7 citations and represents a significant step toward practical deployment of panoptic segmentation in applications like autonomous driving and robotics. By prioritizing both accuracy and efficiency, Weber's work addresses a critical bottleneck in the field, demonstrating that high-quality holistic scene parsing need not come at the cost of computational speed. His research continues to push the boundaries of what is possible in real-time visual perception.