Profile Picture

Arvin Cheng

About Me

fumo

Education & Interests

  • Computer Science Undergraduate @ Pennsylvania State University
  • Interested in: Computer Visions, Computer Graphics, and Computer Architecture

Experience

  • Two years of experiences in Machine Learning, Deep Learning, and Computer Visions
  • Two years of experiences in System Programming
  • Six years of Programming Experience, including C++, C, and Python

Technical Skills

  • Proficient in C++, C, Java, Python, and MATLAB

Currently Learning

  • Computer Hardware, Cuda, OpenGL, Rust

Contact me

Experiences

Perception Team Member / Advanced Vehicle Team
September 2025 - December 2025
System Development Intern / E Ink Corporation
July 2025 - August 2025
Research Assistance / Pennsylvania State University
January 2025 - December 2025

Projects

These are all my main coding projects I have done

Vehicle Vision

Responsible for fused sensor data from LiDAR, Radar, and ZED2i Camera to construct the spatial perception of the vehicle. As a core Perception team member, I developed and maintained the ROS2 system to support vehicle's localization, object detection, and environmental awareness.

vehicle vision

Tennis Balls Anaylsis

Leveraged a ResNet-18 and LSTM architecture to predict ball rotational speed with over 85% accuracy using 3–4 frame image sequences. Applied ray-based geometric analysis to localize the ball's rotation axis, incorporating this information as an auxiliary training signal to improve model performance.

tennis ball

Memory Allocator

Implemented an OS-driven slab allocator with segregated free lists to dynamically manage memory, enabling efficient allocation, deallocation, and reallocation of fixed-size objects.

CPU Prototype

Implemented registers, datapath, control unit, ALU, and other componenets to construct a 32-bit, 5-stage pipelined CPU processor. Recreated 5 main stages of MIPS processor to handle data hazards and data forwarding to ensure pipeline integrity.

cpu waveform

Pinhole Camera Model

Utilized the Pinhole Camera Model to map 3D world points to 2D image coordinates using intrinsic and extrinsic camera parameters from two cameras, enabling accurate geometric reasoning for vision tasks.

Pinhole camera model