cv

Ph.D. candidate

The University of Texas at Austin, United States

Bachelor of Engineering

University of Electronic Science and Technology of China, China

Research Scientist Intern

Meta Reality Lab

• Proposed and studied Multimodal Generalization (MMG), a novel and practical problem to investigate how a multimodal system can generalize when data from certain modalities is limited or missing.
• Introduced MMG-Ego4D, a dataset to facilitate the study of MMG problem in ego-centric action recognition task, under both many-shot and few-shot settings. Built a strong baseline model to solve MMG problem, using contrastive learning and cross-modality alignment.

Research Intern

PicsArt AI

• Delivered a few-shot generative adversarial network, which can learn new image classes with minimum computational cost incrementally. Designed a hypernetwork to enable the efficient new classes learning ability, improved model’s generalizability via weakening the discriminator and involving data augmentation.

Research Intern

Facebook AI

• Designed a high-performance incremental few-shot object detection model. Proposed a weakly-supervised data augmentation technique and a compact architecture design to improve the model’s generalizability.

Research Intern

Facebook AI

• Designed an efficiency-orientated neural architecture search algorithm for video action recognition task. Proposed a multivariate two-stream search space and a progressive search strategy.

Research Intern

Horizon Robotics

• Proposed a neural architecture search algorithm for pose estimation task. Designed an effective multi-scale search space and a bi-level search algorithm for macro structure-wise and micro cell-wise search.