Research Interests

Exploring the frontiers of computer vision and machine learning to build intelligent systems that understand and interact with the world around us.

Back to Home

Current Research Focus

Universal Representation Learning

Building models that generalize across multiple tasks and domains with unified representations.

Cross-domain Few-shot Meta-learning

Multi-Task Learning

Creating efficient systems that excel at multiple visual understanding tasks simultaneously.

Dense Prediction Partial Annotation Task Balancing

Limited Supervision Learning

Enabling visual models to learn effectively with minimal human annotation.

Semi-supervised Self-supervised Weakly-supervised

3D-Aware Computer Vision

Building systems that understand three-dimensional structure and spatial relationships.

Depth Estimation 3D Reconstruction Spatial Reasoning

Generative Modeling & Motion

Advancing realistic visual content and human motion synthesis with multimodal approaches.

Motion Generation Gaze-guided Diffusion Models

Methodological Foundations

Developing fundamental algorithmic and theoretical contributions for ML and computer vision.

Optimization Regularization Interpretability