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..
UltraLED: Learning to See Everything in Ultra-High Dynamic Range Scenes
Authors: Yuang Meng, Xin Jin, Lina Lei, Chun-Le Guo, Chongyi Li
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that Ultra LED significantly outperforms existing single-frame approaches. Our code and dataset are made publicly available at https://srameo.github.io/projects/ultraled. ... We conduct both quantitative and qualitative evaluations on the UHDR dataset. Additionally, we observed that some low-light datasets, such as the SID dataset [10], often unintentionally capture UHDR scenes during outdoor night photography. ... In this section, we provide ablation studies to demonstrate the effectiveness of our pipeline and different modules. |
| Researcher Affiliation | Academia | Yuang Meng Xin Jin Lina Lei Chun-Le Guo Chongyi Li VCIP, CS, Nankai University |
| Pseudocode | No | The paper describes its methodology using text, equations, and diagrams (e.g., Figure 2 for framework overview), but it does not contain any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code and dataset are made publicly available at https://srameo.github.io/projects/ultraled. |
| Open Datasets | Yes | Our code and dataset are made publicly available at https://srameo.github.io/projects/ultraled. ... To systematically assess the reconstruction performance of various methods on UHDR scenes, we construct a UHDR image dataset comprising RAW images and their corresponding RGB counterparts. |
| Dataset Splits | No | To systematically assess the reconstruction performance of various methods on UHDR scenes, we construct a UHDR image dataset comprising RAW images and their corresponding RGB counterparts. ... In total, the dataset includes 585 paired RAW images for evaluation. ... We train our network using the synthesized data based on the ground truth of the Raw HDR dataset [57]... |
| Hardware Specification | Yes | In our implementations, we used a single NVIDIA Ge Force RTX 3090 GPU (24 GB), paired with a 20-core CPU and 64 GB of RAM for training and testing. |
| Software Dependencies | Yes | The Torch version is 1.13.1, and the CUDA version is 12.2. |
| Experiment Setup | Yes | Our ratio map estimator employs a simple U-Net architecture [44]... The hyperparameter σ used to construct the ratio map encoding is set to 30 in our experiments. ... The ratio map estimator requires relatively fewer training iterations, only 30,000 iterations using the Adam optimizer, with a learning rate of 5 10 5. |