DF3DV-1K
DF3DV-1K is a large-scale, diverse, distractor-free dataset that provides both cluttered and clean images for each scene.
DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesis
ECCV 2026
1 University of Technology Sydney, 2 University of Sydney, 3 National Yang Ming Chiao Tung University
* Equal Contribution
Abstract
Advances in radiance fields have enabled photorealistic novel view synthesis. In several domains, large-scale real-world datasets have been developed to support comprehensive benchmarking and to facilitate progress beyond scene-specific reconstruction. However, for distractor-free radiance fields, a large-scale dataset with clean and cluttered images per scene remains lacking, limiting the development. To address this gap, we introduce DF3DV-1K, a large-scale real-world dataset comprising 1,048 scenes, each providing clean and cluttered image sets for benchmarking. In total, the dataset contains 89,924 images captured using consumer cameras to mimic casual capture, spanning 128 distractor types and 161 scene themes across indoor and outdoor environments. A curated subset of 41 scenes, DF3DV-41, is systematically designed to evaluate the robustness of distractor-free radiance field methods under challenging scenarios. Using DF3DV-1K, we benchmark nine recent distractor-free radiance field methods and 3D Gaussian Splatting, identifying the most robust methods and the most challenging scenarios. Beyond benchmarking, we demonstrate an application of DF3DV-1K by fine-tuning a diffusion-based 2D enhancer to improve radiance field methods, achieving average improvements of 0.96 dB PSNR and 0.057 LPIPS on the held-out set (e.g., DF3DV-41) and the On-the-go dataset. We hope DF3DV-1K facilitates the development of distractor-free vision and promotes progress beyond scene-specific approaches.
Leaderboard
Dataset
Rank by
| Rank | Method | PSNR ↑ | SSIM ↑ | LPIPS ↓ | Paper | Code |
|---|---|---|---|---|---|---|
| Loading leaderboard... | ||||||
DF3DV-41
We systematically design several challenging scenarios to facilitate and stress-test the development of radiance fields. This also serves as a “zoom-in-no-more” visual benchmark.
DI2FIX
Going beyond scene-specific methods, we introduce DI2FIX (Distractor-Free DIFIX). Although not perfect, DI2FIX can largely mitigate the impact of distractor artifacts and improve rendering quality even for the vanilla 3DGS.
Team
DF3DV-1K Team Members
Yi-Shan Hung
Wei-Ling Chi
Hao-Ping Wang
Charlie Li-Ting Tsai
Yu-Cheng Chang
Thomas Do
Honorary Members
Fu-Jung Liu
Li-Chen Liu
Su-Lan Liu
Yu-Jun Huang
BibTeX
@article{lu2026df3dv,
title={DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesis},
author={Lu, Cheng-You and Hung, Yi-Shan and Chi, Wei-Ling and Wang, Hao-Ping and Tsai, Charlie Li-Ting and Chang, Yu-Cheng and Liu, Yu-Lun and Do, Thomas and Lin, Chin-Teng},
journal={arXiv preprint arXiv:2604.13416},
year={2026}
}