EN|ZH

SLAM · State Estimation · Autonomous Driving

Mapping, localizing,
learning, writing.

I'm Yu Zhang. I work on autonomous driving algorithms, going deep in SLAM and robotics. I enjoy exploring all kinds of things, and I like writing down how I grow and think along the way.

More about me →

fig. pose graph, 42 keyframes

Featured

pinned
GAVIS overview: anisotropic visibility field quantifies uncertainty by modeling which regions are observed by training views. Left room (visible) → low uncertainty; right room (invisible) → high uncertainty in the rendered uncertainty map.

fig. GAVIS overview: anisotropic visibility field quantifies uncertainty by modeling which regions are observed by training views. Left room (visible) → low uncertainty; right room (invisible) → high uncertainty in the rendered uncertainty map.

K2E-B-G2-7 · PAPER NOTE

GAVIS: Anisotropic Visibility Field for Uncertainty-Driven 3DGS Active Mapping

GAVIS paper note — uncertainty quantification for 3DGS via a per-particle anisotropic visibility field; spherical harmonics representation, 200+ FPS real-time UQ, outperforms FisherRF/VIMC/NVF across all image-quality metrics

2026-07-13 · slam / papers

Uncertainty maps for an unconverged vs. converged 3DGS scene: rendered image, render error, and estimated color/semantic uncertainty. Estimated uncertainty closely tracks true render error in both regimes.

fig. Uncertainty maps for an unconverged vs. converged 3DGS scene: rendered image, render error, and estimated color/semantic uncertainty. Estimated uncertainty closely tracks true render error in both regimes.

K2E-B-G6-6

These Magic Moments: Differentiable Uncertainty Quantification of Radiance Field Models

2026-07-13 · slam / papers

Per-pixel uncertainty estimation for novel view synthesis: ground truth, RGB render, predicted uncertainty map, and true error (DSSIM). The post-hoc uncertainty map closely mirrors regions of actual rendering error.

fig. Per-pixel uncertainty estimation for novel view synthesis: ground truth, RGB render, predicted uncertainty map, and true error (DSSIM). The post-hoc uncertainty map closely mirrors regions of actual rendering error.

K2E-B-G6-7

3DGS-U: Predictive Photometric Uncertainty in Gaussian Splatting

2026-07-13 · slam / papers

Recent Notes

220 篇 · View all →
slam 83programming-language-stack 82system-design 28operating-system 21math 2robot 2foundations 1hardware 1

Thoughts

View all →

Reflections on career, technology, and engineering culture.