🥳

Linji (Joey) Wang

PhD Student in AI & Robotics

CS PhD at George Mason University

MechE MS at Carnegie Mellon University

MechE BS at University of Cincinnati

Short Bio

Linji (Joey) Wang is a PhD student in Computer Science at George Mason University, focusing on AI and Robotics. With a background in Mechanical Engineering from Carnegie Mellon University and the University of Cincinnati, Joey’s research explores the intersection of deep learning and robotics. His current work on Grounded Curriculum Learning demonstrates his commitment to enhancing real-world reinforcement learning in robotics through innovative AI techniques. Joey aims to leverage generative AI to bridge the gap between simulated and real-world task distributions, ultimately creating more adaptable and efficient robotic systems.

Experience

Research Assistant

Robotixx Lab, George Mason University

September 2023 – May 2024 Fairfax, VA

Grounded Curriculum Learning for Efficient Reinforcement Learning in Robotics

Teaching Assistant

George Mason University

September 2023 – May 2024 Fairfax, VA

Introduction to Programming

Teaching Assistant

Carnegie Mellon University

September 2022 – Present Pittsburgh

Machine Learning
Deep Learning

Research Assistant

Computational Engineering & Robotics Laboratory

January 2022 – Present Pittsburgh

Computer Vision
Augmented Reality
3D Scene Understanding

Education

PhD in Computer Science

George Mason University

September 2023 – May 2027 Farifax, VA

MS in Mechanical Engineering

Carnegie Mellon University

September 2021 – May 2023 Pittsburgh, PA

BS in Mechanical Engineering

University of Cincinnati

September 2016 – May 2021 Cincinnati, OH

BS in Mechanical Engineering

Chongqing University

September 2016 – May 2021 Chongqing, China

Recent Posts

Colorizing the Prokudin-Gorskii Photo Collection

This project requires you to use image processing techniques to create a color image from the digitized Prokudin-Gorskii glass plate photographs, with the goal of producing an image with as few visual artifacts as possible. A digital picture of a glass plate, with its three channels arranged from top to bottom as BGR, serves as the process’s input. Our task is to isolate the three channels and properly align them.

Linji (Joey) Wang

Last updated on May 9, 2023 5 min read Project

Colorizing the Prokudin-Gorskii Photo Collection

Explore Test 2 Image Generating

In this project, we first attempted to reproduce and explore the GigaGAN model, uncovering its remarkable text-to-image processing techniques. Subsequently, we shifted our focus to the MinImagen2 architecture, an efficient text-to-image generation model capable of producing high-quality images. Our goal is to experiment with these two models, comparing their strengths and weaknesses, and ultimately share our findings and experiences in the world of image generation.

Linji (Joey) Wang

Last updated on May 9, 2023 7 min read Project

GAN Photo Editing

In this project, we explore the applications of Generative Adversarial Networks (GANs) for photo editing tasks such as image reconstruction, interpolation, and synthesis.

Linji (Joey) Wang

Last updated on Apr 11, 2023 5 min read Project

Gradient Domain Fusion

Welcome to our website about Gradient Domain Fusion, a powerful technique that allows for seamless merging of multiple images into a single high-quality output. Our project aims to explore this technique and provide a detailed guide on how to implement it effectively.

Linji (Joey) Wang

Last updated on May 9, 2023 7 min read Project

Neural Style Transfer

I’m excited to share my recent project on Neural Style Transfer, which is a fascinating technique that combines the content of one image with the artistic style of another, resulting in a stunning and unique blend of the two.

Linji (Joey) Wang

Last updated on Apr 3, 2023 5 min read Project

See all posts

Projects

Grounded Curriculum Learning for Robotics

We introduce Grounded Curriculum Learning (GCL), a novel framework that improves real-world reinforcement learning in robotics by aligning simulated task distributions with real-world tasks while considering task sequences and robot performance history.

When Cats meet GANs

In this assignment, we implemented two types of GANs - a Deep Convolutional GAN (DCGAN) and a CycleGAN. The DCGAN was trained to generate grumpy cats from random noise, while the CycleGAN was trained to convert between two types of cats (Grumpy and Russian Blue) and between apples and oranges. Both GANs were implemented with data augmentation and differentiable augmentation techniques.