Jingxi Yu

Knowledge Background

Main Research Focus: Image Processing, Computer Vision, Machine Learning

Specialized Area: Object Detection and Tracking

Programming Background: Python, Java (Proficient); C, C# (Intermediate)

Education:

Sept 2016 – June 2020: Department of Mathematics, University of Washington (Bachelor of Arts)

Sept 2020 - Present: Department of Electrical & Computer Engineering, University of Washington (Master of Science)

Research Projects

Developed and now maintaining a video annotation graphical user interface for object detection and tracking in a project with National Oceanic and Atmospheric Administration (NOAA). This GUI is a semi-automatic system that reads in fish tracking prediction results from deep learning algorithms and allows human annotator to adjust them. The GUI is developed using C# with the capability of reading in more than 50000 images per haul.

Implemented a GUI that can perform image segmentation using three different methods: Otsu, K-means, and EM-Gaussian Mixture Models. The intention is to show the results of image segmentation obtained from each of the three methods, while allowing user to change the k-value to see the different image outputs.

Currently working on fish detection and classification using MMDection based on video data captured from underwater cameras. This is a collaborative project with Alaska Fisheries Science Center at NOAA. The current best Mean Average Precision (mAP) is obtained by using the RetinaNet model and RetinaNetResNeXt-101 as backbone. The dataset was accepted as a workshop paper to the Eight Workshop on Fine-Grained Visual Categorization (FGVC8) at the Computer Vision and Pattern Recognition Conference 2021 (CVPR 2021).

Also participating in the 2nd Anti-UAV Workshop & Challenge at International Conference on Computer Vision 2021 (ICCV 2021). Provide with 140 high quality, Full HD video sequences, the goal is to detect and track the UAV within the video. Have tried using YOLO v3, Faster RCNN and RetinaNet while creating a new model based on Feature Pyramid Networks (FPN) and CenterNet. The current best result at AP50 is around 90% and still has improving capability.

Internship

Currently doing an internship at Ernst Leitz Labs under Leica Camera AG. Working as in Imaging Engineer and doing research on training an machine learning model to replace human mobile camera ISP tunning.

Publications

SEAMAPD21: a large-scale reef fish dataset for fine-grained categorization. The Eight Workshop on Fine-Grained Visual Categorization (FGVC8). June 2021

Design and Implementation of an Android Mobile Phone Guide System Based on Cloud Platform. Electronic Technology and Software Engineering. 2015, (66), pp. 88-90 (sole author)

Network Selection Algorithm for Multiservice Multimode Terminals in Heterogeneous Wireless Networks. IEEE Access, 2019, vol. 7, pp. 46240-46260 (third author)

Network Selection Algorithm Based on Chi-Square Distance in Heterogeneous Wireless Networks. Wireless Personal Commun. 2019, https://doi.org/10.1007/s11277-019-06946-2. pp. 1-19 (third author)

Enhanced Correlation Filter with Reinforcement Learning and Object Motion Model for Visual Tracking. The IEEE International Conference on Multimedia & Expo (ICME). 2020. Submitted. (third author)

Guide System based on Android Mobile Phone. Computer Software Copyright, 2015

Contact

• Twitter
• Facebook
• Instagram
• GitHub

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';