Zhao (Joey) Zhang （张 钊）

I'm currently working as a Staff Research Scientist at Canva Canva , with a focus on Image Generation and Multimodal LLMs. I completed my Master's degree at Nankai University, where I was under the supervision of Ming-Ming Cheng. Please feel free to contact me at (📮: zzhang🥳mail🔅nankai🔅edu🔅cn)

You can also find me on Google Scholar profile of Zhao Zhang and Bilibili profile of Zhao Zhang

Duck illustration with a camera

Recent News

  • [🔝/2026] Canva We're seeking interns passionate about Image Generation / MLLM to collaborate on impactful research projects. Please feel free to contact me via email.
  • [10/2026] 🔥 We released Canva Craft, Canva's first image generation model trained from scratch, placing Canva Research #2 among AI labs on DesignArena's image generation leaderboard, behind only OpenAI.
  • [10/2026] 🎉 Two papers, DAD and PixelART, were accepted to NeurIPS 2026.
  • [03/2026] Magic Layers Magic Layers is now live on Canva, turning flat or AI-generated images into fully editable, multi-layer designs.
  • [02/2026] Two papers were accepted by CVPR 2026. See you in Denver, Colorado!
  • [06/2025] We released CreatiPoster , an AI-driven graphic design generation system for multi-layer and editable compositions with strong visual appeal.
  • [05/2025] Our Unified-MLLM for image layer decomposition was accepted by ICML 2025.
  • [01/2025] RelationLMM was accepted by TPAMI-25.
  • [12.2024] Our GRES is selected as an "Excellent Science & Technology Academic Paper" for the 2024 Shenzhen 4th Excellent Science & Technology Academic Paper Selection.
  • [04/2024] 🔥 Graphist was accepted by AAAI 2025. We have unleashed the potential of MLLM in graphic design.
  • [02/2024] One paper was accepted by CVPR 2024.
  • [09/2023] D-Cube was accepted by NeurIPS 2023.
  • [08/2023] We released Link-Context Learning for MLLMs as well as an interesting dataset ISEKAI.
  • [07/2023] GRES was accepted by ICCV 2023.
  • [06/2023] 🔥 We released Shikra , an awesome MLLM for Referential Dialogue.
Show more news Show less news
  • [01/2023] One paper was accepted by TPAMI.
  • [08/2022] One paper was accepted by ECCV 2022.
  • [07/2022] Two papers were accepted by ACM MM 2022.
  • [07/2022] I am working as a vision-language researcher in SenseTime Research.
  • [06/2022] One paper was accepted by CVMJ.
  • [03/2022] One paper was accepted by CVPR 2022 as an oral presentation.
  • [12/2020] I am working as an intern in Tencent Youtu Lab.
  • [12/2020] One paper was accepted by TIP 2021.
  • [07/2020] One paper was accepted by ECCV 2020.
  • [05/2020] One paper was accepted by TNNLS 2020.
  • [04/2020] One paper was accepted by CVPR 2020.
  • [09/2019] I have joined the Media Computing Lab under the supervision of Prof. Ming-Ming Cheng!
  • [06/2019] I graduated from Yangzhou University, and received my bachelor degree.

Experiences

  • Staff Research Scientist
    2025 - Now
    Canva Original Research Exploration (CORE)
    Canva
Canva
  • Expert Researcher in Vision & Language
    2023 - 2025
    Intelligent Creation
    ByteDance
ByteDance
  • Researcher in Vision & Language
    2022 - 2023
    Smart City Group (SCG)
    SenseTime
SenseTime
  • Internship in Computer Vision
    2020 - 2021
    Youtu Lab
    CSIG, Tencent
Tencent Youtu Lab
  • M.S. in Computer Science
    2019 - 2022
    Media Computing Lab (supervised by Prof Ming-Ming Cheng)
    School of Computer Science, Nankai University
Nankai University
  • B.S. in Computer Science
    2015 - 2019
    College of Innovation and Entrepreneurship (Elite College)
    School of Information Engineering, Yangzhou University
Yangzhou University

Applications

Canva Craft

Now live on

Canva Craft is Canva Research's first image generation foundation model trained from scratch. Its two-stage architecture plans composition and typography before rendering images. At launch, Craft placed Canva Research second among AI labs on DesignArena's image generation leaderboard, behind only OpenAI.

[Try Craft] [Research] [News]

Magic Layers

Now live on

Magic Layers turns flat or AI-generated images into fully editable, multi-layer designs, enabling users to modify text, objects, and layouts without regenerating the whole image.

Related paper: Masked Region Transformer, CVPR 2026.

[Try Magic Layers] [News] [YouTube]

Layout to Design

Available on

Layout to Design generates complete and visually appealing designs from user-provided text, assets, and an initial layout, helping users quickly create polished posters, product displays, and promotional content.

Related papers: CreatiPoster and CreatiDesign.

[News]

Layout to Design preview

Selected Publications

  • Detect Anything in Graphic Design:
    Element-Level Rewards for Autoregressive Detection
    Jiangning Zhu, Bowen Li, Shenyu Qiao, Yima Gu, Zhao Zhang, Yuhui Yuan, Shixia Liu
    NeurIPS 2026   [Paper] [PDF] [Code] [bib]

DAD concept illustration: element detection boxes around typography, fruit, and a leaf in a graphic design

  • PixelART: Image-to-Layer Decomposition
    without Latents or Text-to-Image Pretraining
    Zelin Jia, Zhao Zhang, Zhicong Tang, Yuhui Yuan, Shixia Liu
    NeurIPS 2026   [Paper] [PDF] [bib]

PixelART concept illustration: a floral poster decomposed into background, flowers, and transparent typography layers

  • Magic Layers logo Masked Region Transformer for Layered Image
    Generation and Editing at Scale
    Zhicong Tang, Jingye Chen, Zhao Zhang, Mohan Zhou, Yuchi Liu,
    Yifan Pu, Yalong Bai, Ethan Smith, Yuhui Yuan
    CVPR 2026   [Paper] [Try Magic Layers]

Visual overview of Masked Region Transformer for Layered Image Generation and Editing at Scale

  • Layton: Latent Consistency Tokenizer
    for 1024-pixel Image Reconstruction and Generation by 256 Tokens
    Qingsong Xie, Zhao Zhang, Zhe Huang, Yanhao Zhang, Haonan Lu, Zhenyu Yang
    CVPR 2026   [PDF] [Project] [bib]

Visual overview of Layton: Latent Consistency Tokenizer for 1024-pixel Image Reconstruction and Generation by 256 Tokens

  • CreatiPoster: Towards Editable and Controllable
    Multi-Layer Graphic Design Generation
    Zhao Zhang, Yutao Cheng, Dexiang Hong, Maoke Yang,
    Gonglei Shi, Lei Ma, Hui Zhang, Jie Shao, Xinglong Wu
    arXiv 2025   [Repo] [Paper]

Visual overview of CreatiPoster: Towards Editable and Controllable Multi-Layer Graphic Design Generation

  • CreatiDesign: A Unified Multi-Conditional Diffusion Transformer
    for Creative Graphic Design
    Hui Zhang, Dexiang Hong, Maoke Yang, Yutao Cheng, Zhao Zhang, Weidong Chen,
    Jie Shao, Xinglong Wu, Zuxuan Wu, and Yu-Gang Jiang
    ICLR 2026   [Repo] [Project page] [Paper] [bib]

Visual overview of CreatiDesign: A Unified Multi-Conditional Diffusion Transformer for Creative Graphic Design

  • Decomposition of Graphic Design with Unified Multimodal Model
    Hui Nie, Zhao Zhang, Yutao Cheng, Maoke Yang, Gonglei Shi, Qingsong Xie, Jie Shao, Xinglong Wu
    ICML 2025   [Repo Coming Soon]

Visual overview of Decomposition of Graphic Design with Unified Multimodal Model

  • RelationLMM: Large Multimodal Model
    as Open and Versatile Visual Relationship Generalist
    Chi Xie, Shuang Liang, Jie Li, Zhao Zhang, Feng Zhu, Rui Zhao, Yichen Wei
    TPAMI 2025   [Paper] [bib]

Visual overview of RelationLMM: Large Multimodal Model as Open and Versatile Visual Relationship Generalist

  • Graphic Design with Large Multimodal Model
    Yutao Cheng*, Zhao Zhang*, Maoke Yang*, Hui Nie, Chunyuan Li, Xinglong Wu, Jie Shao
    AAAI 2025   [PDF] [Project] [bib]

Visual overview of Graphic Design with Large Multimodal Model

  • Link-Context Learning for Multimodal LLMs
    Yan Tai, Weichen Fan, Zhao Zhang, Ziwei Liu
    CVPR 2024   [PDF] [Code] [bib]

Visual overview of Link-Context Learning for Multimodal LLMs

  • Shikra: Unleashing Multimodal LLM’s Referential Dialogue Magic
    Keqin Chen, Zhao Zhang*, Weili Zeng, Richong Zhang, Feng Zhu, Rui Zhao
    arXiv 2023   [PDF] [Code] [bib]

Visual overview of Shikra: Unleashing Multimodal LLM’s Referential Dialogue Magic

  • Described Object Detection: Liberating Object Detection with Flexible Expressions
    Chi Xie*, Zhao Zhang*, Yixuan Wu, Feng Zhu, Rui Zhao, Shuang Liang
    NeurIPS 2023   [PDF] [Code] [bib]

Visual overview of Described Object Detection: Liberating Object Detection with Flexible Expressions

  • Advancing Referring Expression Segmentation Beyond Single Image
    Yixuan Wu*, Zhao Zhang*, Chi Xie, Feng Zhu, Rui Zhao
    ICCV 2023   [PDF] [Code] [bib]

Visual overview of Advancing Referring Expression Segmentation Beyond Single Image

  • User-Oriented Interactive Style Transfer
    Zheng Lin, Zhao Zhang, Kang-Rui Zhang, Bo Ren, Ming-Ming Cheng
    CVMJ 2026   [PDF] [Code] [中译版] [DOI] [bib]

Visual overview of User-Oriented Interactive Style Transfer

  • Image Harmonization by Matching Regional References
    Ziyue Zhu, Zhao Zhang, Zheng Lin, Ruiqi Wu, Zhi Chai, Chun-Le Guo
    arXiv 2022   [PDF] [Code] [中译版] [bib]

Visual overview of Image Harmonization by Matching Regional References

  • Co-Salient Object Detection with Co-Representation Purification
    Ziyue Zhu*, Zhao Zhang*, Zheng Lin, Xing Sun, Ming-Ming Cheng
    TPAMI 2023   [PDF] [Code] [中译版] [bib]

Visual overview of Co-Salient Object Detection with Co-Representation Purification

  • PAC-Net: Highlight Your Video via History Preference Modeling
    Hang Wang, Penghao Zhou, Chong Zhou, Zhao Zhang, Xing Sun
    ECCV 2022   [PDF] [bib]

Visual overview of PAC-Net: Highlight Your Video via History Preference Modeling

  • Multi-Mode Interactive Image Segmentation
    Zheng Lin, Zhao Zhang*, Ling-Hao Han, Shao-Ping Lu
    ACM MM 2022   [PDF] [Code] [中译版]

Visual overview of Multi-Mode Interactive Image Segmentation

  • KnifeCut: Refining Thin Part Segmentation with Cutting Lines
    Zheng Lin, Zheng-Peng Duan, Zhao Zhang, Chunle Guo, Ming-Ming Cheng
    ACM MM 2022 (Oral)   [PDF] [Code] [中译版]

Visual overview of KnifeCut: Refining Thin Part Segmentation with Cutting Lines

  • Sequential Interactive Image Segmentation
    Zheng Lin, Zhao Zhang, Zi-Yue Zhu, Deng-Ping Fan, Xia-Lei Liu
    CVMJ 2022   [PDF] [Code] [中译版]

Visual overview of Sequential Interactive Image Segmentation

  • FocusCut: Diving into a Focus View in Interactive Segmentation
    Zheng Lin, Zheng-Peng Duan, Zhao Zhang, Chun-Le Guo, Ming-Ming Cheng
    CVPR 2022 (Oral)   [PDF] [Code] [中译版] [bib]

Visual overview of FocusCut: Diving into a Focus View in Interactive Segmentation
Visual overview of Gradient-Induced Co-Saliency Detection

  • Bilateral Attention Network for RGB-D Salient Object Detection
    Zhao Zhang, Zheng Lin, Jun Xu, Wenda Jin, Shao-Ping Lu, and Deng-Ping Fan
    TIP 2021   [PDF] [Code] [bib]

Visual overview of Bilateral Attention Network for RGB-D Salient Object Detection

  • Rethinking RGB-D Salient Object Detection: Models, Datasets, and Large-Scale Benchmarks
    Deng-Ping Fan, Zheng Lin, Zhao Zhang, Menglong Zhu, Ming-Ming Cheng
    TNNLS 2020   [PDF] [Code] [Project] [bib]

Visual overview of Rethinking RGB-D Salient Object Detection: Models, Datasets, and Large-Scale Benchmarks

  • Interactive Image Segmentation with First Click Attention
    Zheng Lin, Zhao Zhang, Lin-Zhuo Chen, Ming-Ming Cheng, Shao-Ping Lu
    CVPR 2020   [PDF] [Code] [Project] [bib]

Visual overview of Interactive Image Segmentation with First Click Attention

Services

  • Reviewer for T-PAMI, TIP, TMI, TMM, TCSVT, CVPR, ICCV, ECCV, NeurIPS, EMNLP, ACMMM, etc.