Researcher

Jiaya Jia

Also known as 賈佳亞

Jiaya Jia is a computer vision and AI researcher whose work spans image smoothing, semantic segmentation and multimodal models. He earned a joint computer science PhD from HKUST and Microsoft Research in 2004, joined CUHK as an assistant professor in 2004, became a full professor in 2015 and moved to HKUST in 2024. His work includes the L0 image-smoothing method, ICNet for real-time semantic segmentation and LLaMA-VID, which represents video frames with two tokens. He founded SmartMore in 2019, became an IEEE Fellow in 2018 and an ACM Fellow in 2025, and now directs HKUST's Von Neumann Institute while chairing SmartMore.

Now at
SmartMore Corporation Limited and Hong Kong University of Science and Technology
Works
346
Citations
76,371
h-index
109

Publication numbers from OpenAlex, as of 14 October 2026.

Photo of Jiaya Jia

Timeline

  1. 2003–2010

    A doctorate and CUHK roles

    Jia is a visiting scholar at Microsoft from March 2003 to August 2004 and earns a joint computer science PhD from HKUST and Microsoft Research in 2004. CUHK appoints him assistant professor in 2004 and promotes him to associate professor in 2010. He conducts collaborative research at Adobe Research in 2007.

  2. 2011

    L0 image smoothing

    With Li Xu, Cewu Lu and Yi Xu, Jia coauthors Image Smoothing via L0 Gradient Minimization, an editing method that sharpens major edges and removes low-amplitude structures.

  3. 2015–2018

    ICNet and an IEEE fellowship

    CUHK promotes Jia to full professor in 2015. In 2018, he becomes an IEEE Fellow for contributions to deblurring techniques in computational photography and coauthors ICNet, which uses multiresolution branches and label guidance for real-time semantic segmentation.

  4. 2019

    Founding SmartMore

    Jia founds SmartMore in 2019 to focus on smart-manufacturing optimisation and automation.

  5. 2024

    LLaMA-VID and a texture award

    Jia moves to HKUST as a chair professor in 2024. At ECCV 2024, he coauthors LLaMA-VID, which represents each video frame with a context token and a content token for image and video understanding. At SIGGRAPH Asia 2024, the coauthored paper Structure extraction from texture via relative total variation receives the Test-of-Time Award.

  6. 2025

    An ACM fellowship

    Jia is selected as an ACM Fellow in 2025 for outstanding research achievements in computer science.

  7. As of October 2026

    Directing the Von Neumann Institute

    Jia is a chair professor at HKUST, directs its Von Neumann Institute and is SmartMore's founder and chairman. In September 2026, he presents the institute's work in embodied AI, multimodal reasoning and large AI models.

Sources and credits

Photo: jiaya.me; Jiaya Jia; creator not stated. License: Not stated by source. Source: jiaya.me. Cropped to a square and resized. Ask us to remove or replace this photo.

  1. Jiaya JIA | HKUST CSE cse.hkust.edu.hk. Checked 14 October 2026.
  2. Jiaya Jia – CUHK CSE cse.cuhk.edu.hk. Checked 14 October 2026.
  3. Image Smoothing via L0 Gradient Minimization jiaya.me. Checked 14 October 2026.
  4. Prof. Jiaya Jia Elected IEEE Fellows 2018 – CUHK CSE cse.cuhk.edu.hk. Checked 14 October 2026.
  5. ICNet for Real-Time Semantic Segmentation on High-Resolution Images openaccess.thecvf.com. Checked 14 October 2026.
  6. Soul of sophistication | CUHK in Focus focus.cuhk.edu.hk. Checked 14 October 2026.
  7. LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models ecva.net. Checked 14 October 2026.
  8. SIGGRAPH Asia 2024 Award Winners asia.siggraph.org. Checked 14 October 2026.
  9. Prof. Tim Kwang-Ting CHENG, Prof. Jiaya JIA and Prof. Ke YI Elected as Association for Computing Machinery Fellows 2025 for Outstanding Research Achievements in Computer Science | HKUST CSE cse.hkust.edu.hk. Checked 14 October 2026.
  10. Jiaya Jia zgcforum.com.cn. Checked 14 October 2026.
  11. News | VNI vni.ust.hk. Checked 14 October 2026.