Changan Chen

Papers

Total Citations

H-Index

About

Changan Chen is an emerging researcher whose work centers on computer vision, self-supervised learning, and visual correspondence. His research tackles fundamental challenges in visual understanding, particularly how machines can learn meaningful correspondences across different viewpoints and visual transformations without relying on expensive labeled data. His most notable recent contribution, "Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency" (2025), addresses a critical gap in the field by extending correspondence learning beyond small image transformations to handle challenging cross-view scenarios — a problem deeply connected to how humans perceive and interpret visual scenes. By introducing predictive cycle consistency as a self-supervised training signal, Chen advances the ability of models to establish robust pixel-level mappings across dramatically different viewpoints, pushing the boundaries of what unsupervised visual systems can achieve. Though early in citation accumulation, with his 2025 work already garnering attention, Chen represents a promising voice in the self-supervised learning community. His focus on bridging the gap between human perceptual capabilities and machine vision systems positions him as a researcher worth following as the field of foundation models and embodied AI continues to rapidly evolve.