Research map / 3D vision / Gaussian splatting
DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesis
Cheng-You Lu, Yi-Shan Hung, Wei Chi, Hao Ping Wang, Charlie Tsai, Yu-Cheng Chang, Yu-Lun Liu, Thomas Do, Chin-teng Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune DIFIX on DF3DV-1K* (DI2FIX) as plug-in enhancer | Radiance field renderings · On-the-go + DF3DV-41 | PSNR: 20.82 -> 21.78 | Improved |
| fine-tune DIFIX on DF3DV-1K* (DI2FIX) as plug-in enhancer | Radiance field renderings · On-the-go + DF3DV-41 | LPIPS: 0.211 -> 0.154 | Improved |
| fine-tune DIFIX on DF3DV-1K* (DI2FIX) as plug-in enhancer | Radiance field renderings · On-the-go + DF3DV-41 | SSIM: 0.731 -> 0.737 | No clear effect |
| apply off-the-shelf DIFIX without distractor-free fine-tuning | Radiance field renderings · On-the-go + DF3DV-41 | PSNR: 20.82 -> 20.16 | Hurt |
| fine-tune DIFIX on RobustNeRF instead of DF3DV-1K* (ablation) | DIFIX · On-the-go + DF3DV-41 | PSNR: 20.82 -> 20.54 | Hurt |
| increase training scenes from 250 to 1,007 (ablation) | DI2FIX · On-the-go + DF3DV-41 | PSNR: 21.42 -> 21.78 | Improved |
| use moderate LPIPS filter 0.5 rather than 0.1 or 0.9 (ablation) | DI2FIX · On-the-go + DF3DV-41 | PSNR: 21.55 / 21.73 -> 21.78 | Improved |
| leave-one-method-out evaluation of DI2FIX on unseen radiance field methods (ablation) | DI2FIX · DF3DV-41 / On-the-go | PSNR: worst case -0.1 dB PSNR | No clear effect |
| use newer distractor-free methods (AsymGS) over 3DGS | AsymGS · DF3DV-1K | PSNR: 17.93 -> 20.49 | Improved |
| distractor-free methods on semantically similar distractors, fluid distractors and nighttime scenes | Distractor-free methods · DF3DV-41 scenarios | PSNR gain over 3DGS: AsymGS +1.00, +1.71, +1.37 vs +3.67 on daily scenes | Mixed |
Closest papers by meaning
- DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalization (CVPR 2024 · similarity 0.89)
- DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision (CVPR 2024 · similarity 0.88)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.88)
- 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes (CVPR 2025 · similarity 0.88)
- MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · similarity 0.88)
- NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wild (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
