GARD

GARD: Geometry-Aware Representation Denoising for Robust Multi-view 3D Reconstruction

<img src="https://raw.githubusercontent.com/cvlab-kaist/GARD/main/assets/teaser.png" width="850">

Overview

Feed-forward multi-view 3D reconstruction models perform well under clean, ideal imaging conditions, but degrade sharply on real-world captures corrupted by motion blur, noise, and other artifacts. GARD (Geometry-Aware Representation Denoising) addresses this by performing diffusion-based multi-view restoration directly in the feature space of a feed-forward 3D reconstruction model (Depth Anything 3), rather than restoring pixels first and reconstructing geometry afterward. By denoising geometry-aware feature representations, GARD recovers accurate scene geometry from degraded multi-view inputs; an accompanying RGB decoder then reconstructs high-quality images from the same denoised representations, enabling the simultaneous recovery of 3D scene geometry and high-quality imagery in a single restoration pass.

Full training/inference code, data preparation scripts, and evaluation benchmarks are available at github.com/cvlab-kaist/GARD.

Checkpoints

Usage

git clone https://github.com/cvlab-kaist/GARD.git
cd GARD

uv sync
source .venv/bin/activate

# downloads these checkpoints into ckpts/
bash download_scripts/gard_ckpt/download_ckpts.sh

# evaluate on DA3-BENCH (pose / depth / unposed reconstruction)
bash run_scripts/val/val_GARD_da3_bench.sh

# evaluate on real-world camera motion blur scenes
bash run_scripts/val/val_GARD_real_bench.sh

See the GitHub README for full installation, data preparation, training, and evaluation instructions.

Citation

@article{kim2026geometry,
  title={Geometry-Aware Representation Denoising for Robust Multi-view 3D Reconstruction},
  author={Kim, Jin Hyeon and Lee, Jaeeun and Kim, Claire and Oh, Kyoungjin and Cho, Paul Hyunbin and Min, Jaewon and Choi, Yeji and Park, Jihye and Park, Hyunhee and Park, Minkyu and others},
  journal={arXiv preprint arXiv:2605.26230},
  year={2026}
}

Acknowledgement

We thank the authors of RAE, DepthAnything3, GLD, and Motionblur for their excellent work and code, which served as the foundation for this project.