PocketNeRF
Phone-to-3D indoor reconstruction from a handful of iPhone photos, compressed to sub-8-bit for mobile deployment. Highlighted on the CS231n Spring 2025 reports page.
Python PyTorch
Highlighted project on the Stanford CS231n Spring 2025 reports page.
A lightweight pipeline for 3D indoor reconstruction from a handful of iPhone photos, built on Instant-NGP. Introduces Manhattan-world structural priors for sharper geometry under sparse views, and adversarial content-aware quantization (A-CAQ) to compress models to sub-8-bit precision for mobile deployment. Built for Stanford CS 231N with Aaron Jin and Ryan Suh — I owned the A-CAQ implementation.