Projects

Fitting a 2D neural field and a NeRF from multi-view images, including ray generation, sampling, NeRF architecture, and volume rendering results.

Training UNets: unconditional, time-conditioned, and class-conditioned; results and samples.

Diffusion sampling loops, CFG, image-to-image, visual anagrams, and hybrid images.

Homographies, warping, rectification, feature detection/description, matching, RANSAC, Laplacian blending, and photo mosaics.

Defining correspondences with landmarks, Delaunay triangulation, mid-way face, morph sequence, population mean, caricatures, and extras.

Edges and DoG, unsharp masking, hybrid images, Gaussian/Laplacian stacks, multiresolution blending, and related experiments.

Aligning RGB channels using phase correlation and an image pyramid to colorize Prokudin-Gorskii glass plate photos; results across the dataset.

Probed Florence‑2’s DaViT vision encoder with linear classifiers and trained patch‑level SAEs to discover interpretable sparse dictionary features (SDFs).

Me, online, built with React, Next.js, and Tailwind CSS.

My implementation of a diffusion transformer trained on rectified flow.