MinKyu Lee

Flat minima optimization can transform 3D Gaussian Splatting, enabling it to generalize effectively from sparse views while maintaining high fidelity in novel view synthesis.

Multi-scale GANs don't actually generate images in a coarse-to-fine hierarchy, but a simple alignment loss can fix this, yielding SOTA results with less training.

3D Gaussian Splatting can self-filter inconsistent signals, and this paper shows how to amplify that property to achieve state-of-the-art robustness against distractors without sacrificing performance.

Ditch noisy feature distances: SeaCache uses a spectral filter to cache and reuse intermediate diffusion model outputs, slashing latency while maintaining image quality.