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..
Projection-Manifold Regularized Latent Diffusion for Robust General Image Fusion
Authors: Lei Cao, Hao Helen Zhang, Chunyu Li, Jiayi Ma
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental evidence substantiates that PDFuse achieves highly competitive performance across diverse image fusion tasks. The code is publicly available at https://github.com/Leiii-Cao/PDFuse. 1 Introduction [...] 5 Experiments Configuration. We evaluate our method on three representative image fusion tasks: infrared-visible fusion (IVF), multi-exposure image fusion (MEF), and multi-focus image fusion (MFF). |
| Researcher Affiliation | Academia | 1Electronic Information School, Wuhan University, Wuhan, China 2Suzhou Institute, Wuhan University, Suzhou, China EMAIL, EMAIL |
| Pseudocode | Yes | C Algorithm flowchart This figure presents the algorithmic flowchart of our proposed method, the projection-manifold regularized latent diffusion framework. The flowchart is divided into two variants: the basic version (PDFuse-B) and the enhanced version (PDFuse-E). Algorithm 1 PDFuse-B [...] Algorithm 2 PDFuse-E |
| Open Source Code | Yes | Extensive experimental evidence substantiates that PDFuse achieves highly competitive performance across diverse image fusion tasks. The code is publicly available at https://github.com/Leiii-Cao/PDFuse. |
| Open Datasets | Yes | For IVF, 200 test samples from the LLVIP dataset [15] are used. MEF and MFF are evaluated following the testing setups of the MEFB [62] and MFFW [50] datasets, respectively. In addition, we conducted semantic segmentation experiments on the FMB dataset [22] |
| Dataset Splits | Yes | For IVF, 200 test samples from the LLVIP dataset [15] are used. MEF and MFF are evaluated following the testing setups of the MEFB [62] and MFFW [50] datasets, respectively. In addition, we conducted semantic segmentation experiments on the FMB dataset [22] to verify the effectiveness of the fused images generated by our method in high-level vision tasks. [...] We retrained Seg Ne Xt-B [8] on the fusion outputs of all methods from the FMB training set and evaluated on the corresponding test set [22]; |
| Hardware Specification | Yes | All experiments ran on an NVIDIA Ge Force RTX 3090 GPU and a 2.4 GHz Intel Xeon Silver 4210R CPU. |
| Software Dependencies | No | The paper does not explicitly state specific version numbers for key software components (e.g., Python, PyTorch/TensorFlow versions) used for their implementation. |
| Experiment Setup | No | The paper details the evaluation tasks (IVF, MEF, MFF), datasets used, hardware, and evaluation metrics. It also mentions parameters like λ, γint, γgrad, ω, and ϕ that govern the method's behavior. However, it does not provide concrete numerical values for these specific parameters in the main text or appendices. |