Jinge Ma 马金戈

I am a second-year master's student in Electrical and Computer Engineering at the University of Michigan, specializing in computer vision. I am honored to work with Shuang Li and Yilun Du since the summer of 2022, under the supervision of Professor Joshua Tenenbaum. I also collaborate with Zhaoying Pan and Yutong Xie on an explorative project, advised by Prof. Qiaozhu Mei. Before my master's program, I received my Bachelor of Engineering in Electronic and Information Engineering, from University of Chinese Academy of Sciences. During my bachelor's study, I conducted research at the Pattern Recognition and Intelligent System Development Laboratory (PriSDL), with Xiaosong Zhang as my collaborator and under the supervision of Professor Qixiang Ye . I am currently applying for Ph.D. positions of 2023 Fall!

Email: jingema@umich.edu

CV Email Github LinkedIn

Profile picture


University of Michigan Sep.2021 - May.2023(expected)
Master of Science in Electrical and Computer Engineering
Specialization: Computer Vision
GPA: 4.0/4.0
Advisor: Prof. Qiaozhu Mei
Selected Courses:
University of Chinese Academy of Sciences Sep.2016 - July.2020
Bachelor of Engineering in Electronic and Information Engineering
GPA: 3.40/4.0
Thesis: FreeAnchor algorithm based SSD detector. (Chinese National Patent ZL202010778936.2)
Advisor: Prof. Qixiang Ye

Current Research

Project image
Compositional Diffusion Model
Collaborator: Shuang Li and Yilun Du. Supervisor: Prof. Joshua Tenenbaum
  • Grounded language understanding with GLIDE
  • Decomposed objects and attributes in the scene on CLEVR and CelebA dataset
  • Designed a text encoder to learn to the correlation between objects in texts and images.
  • This image shows how the diffusion model decomposes different facial attributes and composes them into a single image, which is conditioned on all the attributes. Zoom in to view it more clearly.


Project image
A Prompt Log Analysis of Text-to-Image Generation Systems
Yutong Xie*, Zhaoying Pan, Jinge Ma**, Luo Jie, Qiaozhu Mei. (equal contribution)
The ACM Web Conference (WWW), The Creative Web Track, 2023
Paper / Code /
            title={A Prompt Log Analysis of Text-to-Image Generation Systems},
            author={Xie, Yutong and Pan, Zhaoying and Ma, Jinge and Luo, Jie and Mei, Qiaozhu},
            journal={arXiv preprint arXiv:2303.04587},

We explored the large-scale prompt logs collected from multiple text-to-image generation systems to investigate the information needs, and provided concrete implications on how to improve text-to-image generation systems for creation purposes.
Project image
Face Animation with Multiple Source Images.
Jinge Ma*, Zhaoying Pan* (equal contribution)
arXiv, 2022
Paper / Colab /
        title={Face Animation with Multiple Source Images},
        author={Ma, Jinge and Pan, Zhaoying},
        journal={arXiv preprint arXiv:2212.00256},

We proposed a flexible schema to enhance the performance of existing face-animation methods by enabling multiple source images as inputs.
Project image
Realization Scheme for Visual Cryptography with Computer-generated Holograms.
Tao Yu, Jinge Ma, Guilin Li, Dongyu Yang, Rui Ma, and Yishi Shi.
International Workshop on Holography and related technologies, 2018
        title={Realization Scheme for Visual Cryptography with Computer-generated Holograms},
        author={Yu, Tao, Ma, Jinge and Li,  Guilin et al.},
        journal={International Workshop on Holography and related technologies},

We propose to realize visual cryptography in an indirect way, which can expand the application range of visual cryptography and further increase the security.


Project image
Bachelor's Thesis: FreeAnchor algorithm based SSD detector(only Chinese version available)Nov. 2019 - Apr. 2020
Advisor: Prof. Qixiang Ye
  • Back to 2019, CNN-based object detection algorithms utilize the Intersection-over-Union (IoU) metric to determine the allocation of anchors to objects. However, this method has certain limitations. To address this issue, we propose a new training approach called FreeAnchor, which allows for the free allocation of anchors.
  • The main goal of FreeAnchor is to learn the most effective features for classification and object location. It can be used in conjunction with any CNN-based detection method, including Single Shot Detector (SSD).
  • By incorporating FreeAnchor, we are able to improve the accuracy and speed of SSD, resulting in a detector that outperforms traditional single-stage methods, particularly in regards to slender, non-central, and densely-placed objects.
Project image
Object Detection ImplementationAug. 2020 - Oct. 2020
Advisor: Prof. Kun Fu and Prof. Xian Sun
  • I reviewed object detection algorithms, including Faster-RCNN, YOLO v3, and YOLO v4.
  • I implemented YOLO v3 with PyTorch on the DOTA dataset to detect objects in remote sensing images. The image shows one example of object detection on remote sensing images and two zoomed-in parts of the example.
Project image
Image Captioning ImplementationJul. 2019 - Aug. 2019
Advisor: Prof. Kun Fu and Prof. Xian Sun
  • I implemented simple CNN and LSTM with PyTorch and TensorFlow.
  • I Implemented show and tell algorithm on UCM dataset with TensorFlow. The image shows two examples of image captioning on remote sensing images.
Project image
Internship at Institute of Computing TechnologyAug. 2020 - Oct. 2020
  • I studied High-speed interface (JESD204B) and network LTE protocols.
  • I researched and developed video compression algorithm based on JPEG XS.
  • I investigated physical layer signal synchronization, frequency offset estimation, and compensation.
  • I developed the low-latency I2S controller on FPGA.
Project image
Development of portable electrocardiograph based on FPGAMar. 2020 - Jun. 2020
Group Members: Zhaoying Pan, Jiu Chen, Zhihong Zeng, and Qianli Ma
  • I worked as the team leader to complete the design and implementation of the portable electrocardiograph.
  • We implemented the circuit and PCB design, PCB projection and welding, Matlab simulation of signal processing, FPGA-based signal processing, FPGA and WeChat applet-based Bluetooth communication, applet display, and functional interface development as a group.
  • I worked on circuit design, Matlab simulation signal processing, FPGA-based signal processing, and Bluetooth communication.


  • Outstanding Research Group Leader, University of Chinese Academy of Sciences 2019

Homepage Template

Thanks for the website template from Michael Niemeyer. Last updated on 12/31/2022