Principled Probabilistic Imaging using Diffusion Models as Plug-and-Play Priors
Authors: Zihui Wu, Yu Sun, Yifan Chen, Bingliang Zhang, Yisong Yue, Katherine Bouman
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the effectiveness of the proposed method on six inverse problems (three linear and three nonlinear), including a real-world black hole imaging problem. Experimental results indicate that our proposed method offers more accurate reconstructions and posterior estimation compared to existing DM-based imaging inverse methods. |
| Researcher Affiliation | Academia | 1Department of Computing and Mathematical Sciences, Caltech 2Department of Electrical Engineering, Caltech 3Department of Astronomy, Caltech 4Department of Electrical and Computer Engineering, Johns Hopkins University 5Courant Institute of Mathematical Sciences, New York University |
| Pseudocode | Yes | Algorithm 1 Plug-and-Play Diffusion Models (Pn P-DM) |
| Open Source Code | Yes | We have included the code for reproducing the main experimental results of our work in the supplemental material. |
| Open Datasets | Yes | We test our proposed algorithm and several baseline methods on 100 images from the validation set of the FFHQ dataset [38] for five inverse problems: (1) Gaussian deblur with kernel size 61 × 61 and standard deviation 3.0, (2) Motion deblur with kernel size 61 × 61 and intensity of 0.5, (3) Super-resolution with 4× downsampling ratio, (4) the coded diffraction patterns (CDP) reconstruction problem (nonlinear) in [10, 51] (phase retrieval with a phase mask), and (5) the Fourier phase retrieval (nonlinear) with 4× oversampling. |
| Dataset Splits | No | The paper mentions using 'the validation set of the FFHQ dataset [38]' for testing, and refers to pre-trained models. However, it does not specify the training, validation, and test splits used by the authors for their own experiments or for training the models themselves within the scope of this paper. |
| Hardware Specification | Yes | All experiments were performed on NVIDIA RTX A6000 and A100 GPUs. |
| Software Dependencies | No | The paper mentions software components like 'U-Net architecture', 'Big GAN', 'Adam W optimizer', 'deepinv library', and 'ehtim package', but does not provide specific version numbers for these or other key software dependencies. |
| Experiment Setup | Yes | Table 3: List of hyperparameters for the likelihood step of Pn P-DM, Table 4: List of hyperparameters for the annealing schedule of ρ in Pn P-DM. |