Beverly Qin - Portfolio

Hi, I'm Beverly

I design systems that bridge the digital and physical, drawing from my background in computer science and architecture.

Scroll to explore →

Image-to-image translation, inpainting, prompt-guided edits, visual anagrams, and hybrid images with DeepFloyd IF.

A time- and class-conditioned flow-matching UNet trained from scratch to generate MNIST digits.

Generative image experiments with DeepFloyd IF diffusion, denoising, and classifier-free guidance.

Neural fields for 2D image fitting and novel-view synthesis from multi-view images.

A sparse 3D reconstruction pipeline from two images, using SIFT, RANSAC pose estimation, and triangulation.

A computer vision pipeline that detects Harris corners, builds local descriptors, and matches corresponding features between image pairs.

A computer vision pipeline that reconstructs color photographs from three-channel glass plate scans using image alignment and multi-scale search.

A robotic arm fabrication project driven by cellular automata algorithms to generate thread-based images.

A Dynamic Particle Dispersal Simulation for Crowd Flow Optimization in Madrid's Plaza Mayor Christmas Market using C and PCA Analysis

A VR simulation built in Unreal Engine to visualize future transit systems and support Jidu's autonomous vehicle design strategy.

A minishell project in C that replicates core shell functionality with built-in commands, signal handling, and external process execution.

A simple client-server chat program that facilitates group messaging across multiple clients on the system.