Disentangle Nighttime Lens Flares: Self-supervised Generation-based Lens Flare Removal
Authors: Yuwen He, Wei Wang, Wanyu Wu, Kui Jiang
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical evaluations demonstrate the effectiveness of the proposed method in both joint and individual glare removal tasks. Experimental Results Settings. Our method is implemented on Pytorch with an NVIDIA GPU (version RTX 3090) and covers 3000 iterations. Qualitative Evaluation. Quantitative Evaluation. Table 1 illustrates our approach ranks first in both PSNR and SSIM evaluations in joint glow and reflective/ghost flares removal task, in terms of 1.4% and 0.4% advantages compared to SOTAs respectively. Ablation Study |
| Researcher Affiliation | Academia | 1Computer Science and Technology and Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System, Wuhan University of Science and Technology 2School of Computer Science and Technology, Harbin Institute of Technology EMAIL, EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper describes the proposed network architecture and methods using textual descriptions and mathematical equations but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | No | The paper does not contain an explicit statement about releasing source code, a link to a code repository, or mention of code in supplementary materials for the described methodology. |
| Open Datasets | Yes | Our Syn Datasets is stimulated from images in training dataset from Bracket Flare (Dai, Luo et al. 2023) processed with a function γ U(1.4,1.8), since there are no joint flared datasets available. The Reflective Flare Removal task dataset is validated on the (Dai, Luo et al. 2023) benchmark and real world captured dataset. The glow suppression task is validated on ECCV2022 (Jin, Yang, and Tan 2022)(without GT) light effect dataset and real world captured datasets. |
| Dataset Splits | No | The paper mentions using 'Our Syn Datasets' and validating on 'Bracket Flare (Dai, Luo et al. 2023) benchmark' and 'ECCV2022 (Jin, Yang, and Tan 2022) light effect dataset', but it does not specify any particular train/test/validation splits (percentages, counts, or methodology) for these datasets that would allow reproduction of the data partitioning. |
| Hardware Specification | Yes | Our method is implemented on Pytorch with an NVIDIA GPU (version RTX 3090) and covers 3000 iterations. |
| Software Dependencies | No | Our method is implemented on Pytorch with an NVIDIA GPU (version RTX 3090) and covers 3000 iterations. While 'Pytorch' is mentioned, a specific version number is not provided, nor are other software dependencies with their versions. |
| Experiment Setup | Yes | Our method is implemented on Pytorch with an NVIDIA GPU (version RTX 3090) and covers 3000 iterations. For the fuzzy kernel, we sampled z from a uniform distribution from 0 to 1, with a fixed random seed of 0. In particularly, our model uses the LMSE loss in the first 1000 iterations to fit the pixel-level differences between ˆy and y, gk and k. 1-LSSIM loss is added in the latter 2000 iterations to fit the differences in the spatial dimensions of the images. |