--- license: apache-2.0 tags: - super-image - image-super-resolution datasets: - div2k metrics: - pnsr - ssim --- # Multi-scale Residual Network for Image Super-Resolution (MSRN) MSRN model pre-trained on DIV2K (800 images training, augmented to 4000 images, 100 images validation) for 2x, 3x and 4x image super resolution. It was introduced in the paper [Multi-scale Residual Network for Image Super-Resolution](https://openaccess.thecvf.com/content_ECCV_2018/html/Juncheng_Li_Multi-scale_Residual_Network_ECCV_2018_paper.html) by Li et al. (2018) and first released in [this repository](https://github.com/MIVRC/MSRN-PyTorch). The goal of image super resolution is to restore a high resolution (HR) image from a single low resolution (LR) image. The image below shows the ground truth (HR), the bicubic upscaling x2 and model upscaling x2. ![Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 4](images/msrn_4_4_compare.png "Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 4") ## Model description The MSRN model proposes a feature extraction structure called the multi-scale residual block. This module can "adaptively detect image features at different scales" and "exploit the potential features of the image". ## Intended uses & limitations You can use the pre-trained models for upscaling your images 2x, 3x and 4x. You can also use the trainer to train a model on your own dataset. ### How to use The model can be used with the [super_image](https://github.com/eugenesiow/super-image) library: ```bash pip install super-image ``` Here is how to use a pre-trained model to upscale your image: ```python from super_image import MsrnModel, ImageLoader from PIL import Image import requests url = 'https://paperswithcode.com/media/datasets/Set5-0000002728-07a9793f_zA3bDjj.jpg' image = Image.open(requests.get(url, stream=True).raw) model = MsrnModel.from_pretrained('eugenesiow/msrn', scale=2) # scale 2, 3 and 4 models available inputs = ImageLoader.load_image(image) preds = model(inputs) ImageLoader.save_image(preds, './scaled_2x.png') # save the output 2x scaled image to `./scaled_2x.png` ImageLoader.save_compare(inputs, preds, './scaled_2x_compare.png') # save an output comparing the super-image with a bicubic scaling ``` ## Training data The models for 2x, 3x and 4x image super resolution were pretrained on [DIV2K](https://data.vision.ee.ethz.ch/cvl/DIV2K/), a dataset of 800 high-quality (2K resolution) images for training, augmented to 4000 images and uses a dev set of 100 validation images (images numbered 801 to 900). ## Training procedure ### Preprocessing We follow the pre-processing and training method of [Wang et al.](https://arxiv.org/abs/2104.07566). Low Resolution (LR) images are created by using bicubic interpolation as the resizing method to reduce the size of the High Resolution (HR) images by x2, x3 and x4 times. During training, RGB patches with size of 64×64 from the LR input are used together with their corresponding HR patches. Data augmentation is applied to the training set in the pre-processing stage where five images are created from the four corners and center of the original image. The following code provides some helper functions to preprocess the data. ```python from super_image.data import EvalDataset, TrainAugmentDataset, DatasetBuilder DatasetBuilder.prepare( base_path='./DIV2K/DIV2K_train_HR', output_path='./div2k_4x_train.h5', scale=4, do_augmentation=True ) DatasetBuilder.prepare( base_path='./DIV2K/DIV2K_val_HR', output_path='./div2k_4x_val.h5', scale=4, do_augmentation=False ) train_dataset = TrainAugmentDataset('./div2k_4x_train.h5', scale=4) val_dataset = EvalDataset('./div2k_4x_val.h5') ``` ### Pretraining The model was trained on GPU. The training code is provided below: ```python from super_image import Trainer, TrainingArguments, MsrnModel, MsrnConfig training_args = TrainingArguments( output_dir='./results', # output directory num_train_epochs=1000, # total number of training epochs ) config = MsrnConfig( scale=4, # train a model to upscale 4x ) model = MsrnModel(config) trainer = Trainer( model=model, # the instantiated model to be trained args=training_args, # training arguments, defined above train_dataset=train_dataset, # training dataset eval_dataset=val_dataset # evaluation dataset ) trainer.train() ``` ## Evaluation results The evaluation metrics include [PSNR](https://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio#Quality_estimation_with_PSNR) and [SSIM](https://en.wikipedia.org/wiki/Structural_similarity#Algorithm). Evaluation datasets include: - Set5 - [Bevilacqua et al. (2012)](http://people.rennes.inria.fr/Aline.Roumy/results/SR_BMVC12.html) - Set14 - [Zeyde et al. (2010)](https://sites.google.com/site/romanzeyde/research-interests) - BSD100 - [Martin et al. (2001)](https://www.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/) - Urban100 - [Huang et al. (2015)](https://sites.google.com/site/jbhuang0604/publications/struct_sr) The results columns below are represented below as `PSNR/SSIM`. They are compared against a Bicubic baseline. |Dataset |Scale |Bicubic |msrn-bam | |--- |--- |--- |--- | |Set5 |2x |33.64/0.9292 | | |Set5 |3x |30.39/0.8678 | | |Set5 |4x |28.42/0.8101 |**32.19/0.8951** | |Set14 |2x |30.22/0.8683 | | |Set14 |3x |27.53/0.7737 | | |Set14 |4x |25.99/0.7023 |**28.67/0.7833** | |BSD100 |2x |29.55/0.8425 | | |BSD100 |3x |27.20/0.7382 | | |BSD100 |4x |25.96/0.6672 |**27.63/0.7374** | |Urban100 |2x |26.66/0.8408 | | |Urban100 |3x | | | |Urban100 |4x |23.14/0.6573 |**26.12/0.7866** | ![Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 2](images/msrn_2_4_compare.png "Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 2") ## BibTeX entry and citation info ```bibtex @InProceedings{Li_2018_ECCV, author = {Li, Juncheng and Fang, Faming and Mei, Kangfu and Zhang, Guixu}, title = {Multi-scale Residual Network for Image Super-Resolution}, booktitle = {The European Conference on Computer Vision (ECCV)}, month = {September}, year = {2018} } ```