Jingtong Wang

Year in School
Sophomore

Major
Computer Science

Research Interests
Artificial Intelligence

Research Mentor
Prof. Arindam Banerjee

Research/Engagement Experience
My first experience with research was when I worked on a physics research project with a mentor involving astrodynamics and simulating the effects of solar sails on spacecraft by creating a mathematical model.

Interests
I enjoy drawing, crocheting, and playing violin.

Project Title
Deep Learning for Remote Sensing

How did I get interested in Computer Science?
I became interested in computer science after taking a course in Java during my sophomore year of high school where I really enjoyed the problem solving aspect of CS and coding games in the class.

What social interests matter to me?
Equity in education

What is my most impactful college experience?
Exploring different clubs on campus.

These are a few of my favorite things!
Listening to music, playing video games, crafting, matcha flavored desserts, and sushi!

Research Description
Remote sensing imagery can provide information that is a crucial part of climate science research and has many societal applications such as resource management and natural disaster monitoring. By integrating remote sensing imagery with deep learning, many traditional tasks involving geospatial data and remote sensing can be modeled more effectively than shallow representations. With the development of TorchGeo, an open source python library in the PyTorch environment that integrates geospatial data with deep learning, we are able to explore machine learning models of the data as well as its effectiveness, opening more paths to possible machine learning solutions. With the library, we first worked on understanding the advantage of prediction accuracy increase with models finetuned on a pre-trained generic model over models with randomized initialization, and currently, we are focusing on comparing the effectiveness of pre-trained models over different datasets (satellite dataset vs general computer vision dataset) for specific task model training. To explore this, we are pre-training models using TorchGeo with different satellites for tasks such as image classification, as current data and satellite-trained models are limited.

Biography
Jingtong Wang is a sophmore majoring in computer science. She became interested in computer science after taking a course in Java during her sophomore year of high school and seriously considered majoring in it by the end of her junior year. During that time, Jingtong participated in clubs which allowed her to explore different aspects of CS such as web development and worked on physics research with a mentor at SFSU involving astrodynamics and simulating the effects of solar sails on spacecrafts. Through those experiences, she has worked in Java, Javascript, Python, and Html. Outside of computer science, Jingtong also has interests in art, including both drawing and painting. In CS STARS, she is excited to explore and learn more about developments in CS, especially relating to AI.