Rongjun Qin
Papers
2
Total Citations
12
H-Index
2
About
Rongjun Qin is a researcher at the forefront of computer vision and reinforcement learning, with key contributions spanning 3D reconstruction and decision-making under uncertainty. His work in structure-from-motion (SfM) tackles the challenging problem of reconstructing three-dimensional scenes from monocular video in confined, feature-poor environments. In his 2022 paper on monocular video frame optimization for 3D pipe reconstruction (6 citations), he developed a feature-based parallax analysis method that intelligently selects geometrically optimal frames—a critical advance for inspecting narrow drainage pipes and other inaccessible infrastructure. This work addresses a fundamental limitation in SfM: the difficulty of maintaining reconstruction quality in constrained spaces where traditional frame selection fails. Simultaneously, Qin is advancing reinforcement learning through his 2022 paper on adversarial counterfactual environment model learning (6 citations). Here, he introduces a novel framework for building robust environment models that predict action effects, enabling sample-efficient policy learning in domains ranging from robot control to healthcare treatment selection. By leveraging adversarial training to generate counterfactual scenarios, his approach allows agents to safely explore and learn from unlimited simulated trials. Qin’s dual focus on geometric reconstruction and intelligent decision-making positions him as a versatile innovator bridging perception and action.
Research Focus
Key Achievements
Top Papers
- 1
- 2Adversarial Counterfactual Environment Model Learning6 citations · 2022