Dequan Wang

Papers

Total Citations

292

H-Index

About

Dequan Wang is a leading researcher at the intersection of computer vision, robotics, and autonomous systems. His work is defined by a commitment to building robust, interpretable, and practically deployable AI. Wang is best known for his foundational contributions to visual domain adaptation and object-centric learning for robotics. He led the creation of the **VisDA benchmark** (2018, 181 citations), a seminal synthetic-to-real dataset that has become the standard for evaluating how vision models generalize from simulated to real-world environments—a critical challenge for both computer vision and robotic applications. In parallel, Wang pioneered **Deep Object-Centric Policies** for autonomous driving (2019, 103 citations). This work directly addresses the fragility of end-to-end deep learning by proposing models that explicitly represent objects. This approach not only improves robustness in novel scenes but also provides intuitive visualizations, making AI decision-making more transparent and trustworthy for safety-critical tasks like driving. By bridging the gap between synthetic training and real-world deployment, and by championing interpretable architectures, Wang’s research has shaped how modern autonomous systems are designed, earning him recognition as a key innovator in building reliable, real-world AI.