PromptIR: Prompting for All-in-One Image Restoration
Authors: Vaishnav Potlapalli, Syed Waqas Zamir, Salman H. Khan, Fahad Shahbaz Khan
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our comprehensive experiments demonstrate the dynamic adaptation behavior of Prompt IR by achieving state-of-the-art performance on various image restoration tasks, including image denoising, deraining, and dehazing using only a unified Prompt IR model. |
| Researcher Affiliation | Collaboration | Mohamed bin Zayed University of AI, Core42, Linköping University |
| Pseudocode | No | The paper describes the architecture and processes using text and block diagrams (e.g., Figure 3, Figure B.1) but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | Our code and pre-trained models are available here: https://github.com/va1shn9v/Prompt IR. |
| Open Datasets | Yes | For image denoising in the single-task setting, we use a combined set of BSD400 [1] and WED [40] datasets for training. The BSD400 dataset contains 400 training images and the WED dataset has 4,744 images. For single-task image deraining, we use the Rain100L [65] dataset, which consists of 200 clean-rainy image pairs for training, and 100 pairs for testing. Finally, for image dehazing in the single-task setting, we utilize SOTS [31] dataset that contains 72,135 training images and 500 testing images. |
| Dataset Splits | No | The paper specifies training and testing datasets and their sizes, but does not explicitly mention the use or size of a separate validation set or how data was split for validation purposes. |
| Hardware Specification | No | The paper mentions 'The computational resources were provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS)... and by Berzelius resource,' but it does not specify concrete hardware details such as GPU models, CPU types, or memory configurations used for the experiments. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x, CUDA 11.x) that would be needed to reproduce the experiments. |
| Experiment Setup | Yes | The architecture of our Prompt IR consists of a 4-level encoder-decoder, with varying numbers of Transformer blocks at each level, specifically [4, 6, 6, 8] from level-1 to level-4. The total number of prompt components are 5. The model is trained with a batch size of 32 in the all-in-one setting, and with a batch of 8 in the single-task setting. The network is optimized with an L1 loss, and we use Adam optimizer (β1 = 0.9, β2 = 0.999) with learning rate 2e 4 for 200 epochs. During training, we utilize cropped patches of size 128 x 128 as input, and to augment the training data, random horizontal and vertical flips are applied to the input images. |