Jia Li (Iris)

I work as an algorithm engineer in JD.com, Inc. Before that, I receive my master degree from Institute of Information Engineering [IIE], University of Chinese Academy of Sciences [UCAS]. Also I had a great experience of one year as a remote intern in PRADA Lab with Professor Wang and Lijie Hu (2023-2024).

My research spans Computer Vision (CV), AI-Generated Content (AIGC), and recommendation systems, all with an integral focus on real-world human impact and applications. Specifically, I am interested in developing human-centric AI frameworks that enhance user autonomy while addressing ethical challenges. I am deeply motivated to explore how AI technology can safely shape and enrich our future.

Email  /  Scholar ZhiHu

To refine the noisy labels of noisy clients via an end-to-end Optimization framework in the scenario of distributed learning.

In our paper, we propose a novl method called Locate and Forget (LaF), which effectively locates potential target concepts in the image for modification by comparing the syntactic trees of the target prompt and scene descriptions in the input image, intending to forget their existence clues in the generated image.

In this paper, we address the limitations of existing text-to-image diffusion models in generating demographically fair results when given human-related descriptions. To overcome this challenge, we propose Fair Mapping, a general, model-agnostic, and lightweight approach that modifies a pre-trained text-to-image model by controlling the prompt to achieve fair image generation.

In this paper, to enhance the representation space utilization and reduce disparities among different demographic groups, we introduce SA-SVD, a regularization method orthogonal to existing face recognition methods. It has potential to capture the intricate patterns and subtle differences in face features across individuals within certain demographic groups.

Some projects around unity!

I'm the lead programmer for this project. It tells a story about a maid(player)'s journey to unravel the mystery of the tomb's occupant.

relative tags: #Ancient;#History;#2d;#Puzzle;#Plot

We use the movements data from ship monitor including the position information data rather than video in a more lightweight way to recreat the crew action in virtual 3d cockpit software.

relative tags: #ShipMotionReconstruction;

coffee

dance

thanks for the web template.