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..
Diffusion Prior Interpolation for Flexibility Real-World Face Super-Resolution
Authors: Jiarui Yang, Tao Dai, Yufei Zhu, Naiqi Li, Jinmin Li, Shu-Tao Xia
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In extensive experiments conducted on synthetic and real datasets, along with consistency validation in face recognition, DPI demonstrates superiority over SOTA FSR methods. [...] Extensive experiments on both synthetic and real-world datasets demonstrate that our method outperforms SOTA FSR methods. |
| Researcher Affiliation | Academia | 1College of Artificial Intelligence, Nankai University, Tianjin, China 2Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China 3College of Computer Science and Software Engineering, Shenzhen University, China |
| Pseudocode | Yes | Algorithm 1: Diffusion Prior Interpolation, given a diffusion model (ยตฮธ( ), ฮฃฮธ( )) and Corrector CRT( ). |
| Open Source Code | No | The paper does not provide an explicit statement about releasing their own code, nor a link to a code repository for the methodology described. |
| Open Datasets | Yes | For evaluation, we utilize synthetic datasets FFHQ1000 and Celeb A1000 (Liu et al. 2015), along with real-world datasets LFW (Huang et al. 2008), Web Photo (Wang et al. 2021), and WIDER (Yang et al. 2016), serving as our testsets. |
| Dataset Splits | No | The paper mentions using specific datasets as "testsets" (e.g., FFHQ1000, Celeb A1000, LFW, Web Photo, WIDER) and that a pre-trained model was trained on FFHQ 49k. However, it does not provide specific training/validation splits or percentages for the experiments conducted with their proposed DPI method. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU models, CPU types, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions using a pre-trained DDPM from DPS and the Deep Face framework but does not specify version numbers for general software dependencies like Python, PyTorch, or CUDA. |
| Experiment Setup | Yes | For each of these three scales, the parameters (ฯ, s, ฯ) is set to (100, 1.4, 500), (300, 1.2, 750), and (500, 1, 1000) respectively. For real-world datasets, we adhere to the experimental settings in Code Former (Zhou et al. 2022), with fixed parameters set to (500, 1, 1000). The sparsity parameter k for CMs is set to 2 for all experiments. |