FouriDown: Factoring Down-Sampling into Shuffling and Superposing
Authors: Qi Zhu, man zhou, Jie Huang, Naishan Zheng, Hongzhi Gao, Chongyi Li, Yuan Xu, Feng Zhao
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To demonstrate the efficacy of Fouri Down, we conduct extensive experiments on image de-blurring and low-light image enhancement. The results consistently show that Fouri Down can provide significant performance improvements. |
| Researcher Affiliation | Academia | 1University of Science and Technology of China, 2S-Lab, 3Nanyang Technological University, 4Nankai University |
| Pseudocode | Yes | Algorithm 1 Pseudo-code of Fouri Down. Data: Input: x RN C H W . Set of real numbers: R. Set of complex numbers: C. Result: y RN C H |
| Open Source Code | Yes | The code is publicly available to facilitate further exploration and application of Fouri Down at https://github.com/zqcrafts/Fouri Down. |
| Open Datasets | Yes | For image enhancement, we assess our Fouri Down model using the LOL [34] and Huawei [35] benchmarks. The LOL dataset contains 500 image pairs (485 for training, 15 for testing), and the Huawei dataset contains 2480 paired images (2200 for training, 280 for testing). We compare our results with two established baselines, SID [36] and DRBN [37]. For image deblurring, we utilize Deep Deblur [38] and MPRNet [39] on the DVD dataset [40], which includes 2103 training and 1111 test pairs. We further validate our model s generalizability using the HIDE dataset [41]. In the context of image de-noising, our training involves the SIDD dataset [42]. Subsequent performance assessments are carried out on the remaining validation samples from the SIDD dataset and on the DND benchmark dataset [43]. For image dehazing, we employ RESIDE dataset for evaluations. |
| Dataset Splits | Yes | The LOL dataset contains 500 image pairs (485 for training, 15 for testing), and the Huawei dataset contains 2480 paired images (2200 for training, 280 for testing). ... on the DVD dataset [40], which includes 2103 training and 1111 test pairs. ... Subsequent performance assessments are carried out on the remaining validation samples from the SIDD dataset and on the DND benchmark dataset [43]. |
| Hardware Specification | No | The paper mentions 'GPU cluster' but does not provide specific details on CPU, GPU models, or memory used for experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, CUDA). |
| Experiment Setup | No | The paper describes comparison configurations (e.g., replacing down-sampling operators) but does not provide specific hyperparameters such as learning rate, batch size, number of epochs, or optimizer settings for the experiments. |