Lin Bai

• Python, Data Structures and Algorithms
• Flask, Pytorch, Tensorflow, OpenCV, Pandas
• C, Java, SQL, Git, HPC

University of Houston, U.S

Ph.D. Student, Electrical and Computer Engineering • 08/2020 — 08/2025

• Research Assistant in “FARSIGHT” Bio-image and Information Analysis Lab.
• Advisor Dr. Badri Roysam

Northeastern University, China

Master of Science, Computer Application Technology • 09/2017 — 03/2020

• Advisor Pro. Donghong Han

Northeastern University, China

Bachelor of Science, Computer Science $ Technology • 09/2013 — 07/2017

Predict COVID-19 Case Numbers in the Houston Area

• Analyzed and processed real-world data, and combined LSTM and SVR on extracted features to predict future trends.
• Won first place in Datathon Kaggle Competition.
• https://sbmi.uth.edu/datathon/past-events/aug20.htm.

Annotation-Free Nuclei Segmentation in Large-Scale Multiplex Brain Image

• Developed annotation and visualization GUI using Tkinter and Flask, Napari.
• Achieved a SAM-based method benefiting annotation acquisition accurately and automatically.
• Generated a synthetic dataset overcoming the insufficient existence of annotations.
• Proposed a Mask-RCNN-based model, which is able to segment the whole nuclei body from dense areas.

Cross-Domain Sentiment Classification on Amazon Reviews

• Converted sentences into combinations of words via Dependency Syntactic Parsing.
• Constructed a graph by using combinations and sentences as nodes. Edges were created to connect nodes and build bridges between different domains. Trained a semi-supervised Graph Convolutional Network to classify nodes.
• Learned nodes’ knowledge was saved to help the next task under the Lifelong learning framework.

Multi Objects Tracking and Segmentation

• With the public Multi-Objects Tracking dataset, used pre-trained segmentation models to generate and merge instance masks, and trained an appearance encoder to generate features. Performed data association between frames to generate tracklets.
• The method can track and segment objects from offline video.

Identify COVID-19 Lung Diseases

• With 2D X-ray dataset, used pre-traisned U-net to segment lung images. Multiple Deep-learning classification models (VGG16, ResNet50, MobileNet, InceptionV3, AG-CNN) were used to classify lung images.
• Compared all models' performance. Grad-Cam was used to visualize the models by generating an activation map.

Social Network Comments Sentiment Classification

• Labeled Sina Microblog data (five classes), and did text preprocessing. Converted word to vector via Word2Vec and built a classification model via SVM.

Online Shopping Website

• Implemented a shopping website application using SSH with MVC pattern in Java, using MySQL as database, and Tomcat as web server.

Library Management System

• Implemented adding, deleting, updating, and searching book and user information functions in C Language.

• Tutor
• Research Assistant
• Teaching Assistant

Han D, Wei F, Bai L, et al. An Algorithm of Sina Microblog User’s Sentimental Influence Analysis Based on CNN+ ELM Model[C]//Proceedings of ELM 2018 9. Springer International Publishing, 2020: 86-97.