Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Optimize the Unseen - Fast NeRF Cleanup with Free Space Prior
Authors: Leo Segre, Shai Avidan
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our Ne RF cleanup method by analyzing its performance in the Ne RF cleanup task, comparing both quantitative metrics and computational efficiency in terms of cleanup and inference speed. We report numerical results to quantify the reduction of artifacts (floaters) and demonstrate that our approach achieves significant speedup while maintaining high-quality novel view synthesis. |
| Researcher Affiliation | Academia | Leo Segre Shai Avidan Tel Aviv University EMAIL EMAIL |
| Pseudocode | No | The paper describes the method conceptually and visually in Figure 2, but does not present a structured pseudocode or algorithm block. |
| Open Source Code | No | Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [No] Justification: We will provide the code and data preparation and evaluation upon acceptance. |
| Open Datasets | Yes | For quantitative evaluation, we use the Nerfbusters dataset [31], which is the standard benchmark for Ne RF cleanup methods. |
| Dataset Splits | Yes | For quantitative evaluation, we use the Nerfbusters dataset [31], which is the standard benchmark for Ne RF cleanup methods... Nerfbusters uniquely provides evaluation views that are significantly farther from the training cameras... Table 3: Comparison of different methods under sparse and dense view settings. Sparse View # Train Images 24 images (Every 8th) 47 images (Every 4th) 93 images (Every 2nd) 161 images (Regular) |
| Hardware Specification | Yes | All experiments were run on a single NVIDIA RTX A5000 GPU. |
| Software Dependencies | No | All experiments were conducted using the latest version of Nerfacto and Splatfacto from Nerfstudio [30], pre-trained for 30,000 steps. |
| Experiment Setup | Yes | All experiments were conducted using the latest version of Nerfacto and Splatfacto from Nerfstudio [30], pre-trained for 30,000 steps... In our method, density optimization was achieved with 1,000 iterations using 217 randomly sampled points across the 3D scene space. Optimization was carried out through Nerfacto s optimizers... We set λ = 0.1 for all visualizations in the paper... In each iteration, we use 4,096 rays with 352 samples per ray ( 1.45M samples) for reconstruction loss. We chose ( 131K samples), which is approximately one order of magnitude smaller... |