Liya Ji

Liya Ji is a Postdoctoral Fellow at the Hong Kong University of Science and Technology (HKUST), supervised by Prof. Qifeng Chen. Liya Ji obtained a Ph.D degree from HKUST, supervised by Prof. Qifeng Chen and Prof. Qiang Yang. Liya Ji served as Manager, Machine Learning at Lenovo Group Limited from 2018 to 2021. Liya Ji obtained Master of Philosophy at the Department of Computer Science and Engineering from HKUST in 2015 and received her Bachelor of Engineering at the Department of Computer Science and Technology from Xi'an Jiaotong University in 2012. Her research interests include generative models, low-level vision, computational aesthetics, and unified multi-modal generation.

Email  /  CV  /  Bio  /  Scholar  

Education/Experiences

Publications

(* indicates the equal contributions. + indicates the corresponding authors.)

We enhance the realism of the synthetic videos from the simulator CARLA via using the World Foundation Model.

We address the instruction-based image editing with Chain-of-Thought MLLMs reasoning, and diffusion generation

We continually train a pre-trained T2V models via block expansion and LLMs conditioning.

We propose a pixel-wise control framework based on unconditional diffusion models for the degradation-blind inverse problems with the performance of semantic consistency and details preservation.

We propose an image assessment metric presenting human's like based on Instagram dataset.

We propose an automatic framework enabling agile development of detection models and human-in-the-loop annoation.

Academic Services & Awards

Miscellanea