About Me

---

I’m a Master’s student at the University of Pennsylvania with a passion for real-time rendering, game development, and deep learning. During my intern at Thunder Fire Studio, Netease Game, I participated in the Project Mugen with a focus on Unity development.

Portfolio

• (Technical) Art
• Computer Graphics
• Games
• Machine Learning
• Other Projects

---

Unity Grass Rendering

Maya Terrain Authoring Tool

TerrainIt is an authoring tool for generating mountainous terrain mesh.

Co-author: Shutong Wu

Reference: Argudo, Oscar, et al. Orometry-based terrain analysis and synthesis

Lego-ify with Houdini

Jellyfish with Houdini

Kitchen with Maya

---

Cuda Path Tracer

A toy path tracer developed in CUDA.

Artist: NewSee2l035

Vulkan Realtime Ray Tracer

A real-time ray tracer based on Vulkan's ray tracing api and the ReSTIR algorithm. This is the final project for CIS 565 - GPU Programming and Architecture.

CIS 565 Course Projects

Boids Flocking Simulation with CUDA	Grass Rendering with Vulkan
Denoising Path Tracing with CUDA	Real-time Ray Tracing with Vulkan

These are the course projects of CIS 565 - GPU Programming and Architecture. In this course I delved into GPU architecture, CUDA, and Vulkan. Its six non-trivial projects will further develop my C++ programming skills.

Vulkan Realtime Renderer

Naku is a toy rasterization-based renderer developed in Vulkan and C++.

---

Mingle

A platform game demo developed with Unity in the CIGA Game Jam 2023 within 48 hours.

Teleport

A FPS demo allowing players to place portals and fire bullets through them.

Hand Tracking Game

Made a shoot’em up game in C++ and OpenGL. Also, we used YOLO v3 to train a object detection model. The goal of using YOLO was to allow player to control character by waving hands in front of a webcam. This is the project for the Undergraduate Innovation and Entrepreneurship Training Program.

---

Path Tracing Denoising

A multi-scale convolutional neural network enhanced with a separate encoder of G-buffers. It effectively denoises extremely noisy Monte-Carlo rendering and can achieve near real-time speeds.

Publications:

• Monte Carlo Denoising with a Sparse Auxiliary Feature Encoder
  SIGGRAPH Asia ’21 Posters
• Denoising Monte Carlo renderings via a multi-scale featured dual-residual GAN
  The Visual Computer 2021

Generating Anime-style Avatars

I created an anime-style face dataset and trained a FastGan model to generate the avater above. Click to see more.

Custom PyTorch Operators

I used CUDA to develop an Image Warp operator and a Kernel Filter operator for PyTorch, both of which support backpropagation.

---

Discord Chatbot

This is a Discord Chatbot I created using OpenAI's API. You can chat with it and ask it to generate an image.

Anki Tango Master

To help me build Japanese vocabulary, I wrote this Python program that parses a Japanese dictionary and generate HTML word cards that can be imported into Anki.