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..
From Chaos to Clarity: 3DGS in the Dark
Authors: Zhihao Li, Yufei Wang, Alex Kot, Bihan Wen
NeurIPS 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results show that our method outperforms LDR/HDR 3DGS and previous state-of-the-art (SOTA) self-supervised and supervised pre-trained models in both reconstruction quality and inference speed on the Raw Ne RF dataset across a broad range of training views. |
| Researcher Affiliation | Academia | Zhihao Li Yufei Wang Alex Kot Bihan Wen Department of EEE, Nanyang Technology University, Singapore |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | Our code is under internal review and will be released upon paper acceptance. |
| Open Datasets | Yes | We utilize the Raw Ne RF dataset 3, which comprises raw images captured in dark scenes using an i Phone X with various ISO settings. 3https://bmild.github.io/rawnerf/ google-research. Licensed under the Apache License, v2.0. |
| Dataset Splits | Yes | For the full-views training setting, we adhere to the same train-test view splits as in Raw Ne RF. For limited views settings, we randomly select subsets (4, 8, 12, 16, 20 views) from the training views, while maintaining a consistent test view set across all experiments to ensure fair comparison. |
| Hardware Specification | Yes | All 3DGS models are trained on a single NVIDIA RTX 3090 GPU, following the default parameter settings of Scaffold-GS. |
| Software Dependencies | No | The paper mentions building upon 'Scaffold-GS [21]' and pretraining on the 'SID dataset [2]', but does not specify software dependencies with version numbers (e.g., PyTorch, TensorFlow versions, or specific library versions). |
| Experiment Setup | Yes | The hyperparameters ฮปnd and ฮปcov, as defined in Eq. (15) and Eq. (14), are set to 5 and 20 for full-views settings, and 3 and 20 for limited views settings, respectively. All 3DGS models are trained on a single NVIDIA RTX 3090 GPU, following the default parameter settings of Scaffold-GS. |