Eugenesiow

Rank:

Average Model Cost: $0.0000

Number of Runs: 71,272

Models by this creator

edsr-base

edsr-base

eugenesiow

The edsr-base model is a deep learning model designed for image super-resolution tasks. It is based on the Enhanced Deep Super-Resolution (EDSR) architecture, which uses residual learning to improve image quality by upscale low-resolution images. The model is trained on a large dataset of high-resolution images and is capable of producing high-quality super-resolution images with enhanced details and sharpness.

Read more

$-/run

61.1K

Huggingface

pan

pan

Pixel Attention Network (PAN) PAN 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 Efficient Image Super-Resolution Using Pixel Attention by Zhao et al. (2020) and first released in this repository. 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 and model upscaling. Model description The PAN model proposes a a lightweight convolutional neural network for image super resolution. Pixel attention (PA) is similar to channel attention and spatial attention in formulation. PA however produces 3D attention maps instead of a 1D attention vector or a 2D map. This attention scheme introduces fewer additional parameters but generates better SR results. The model is very lightweight with the model being just 260k to 270k parameters (~1mb). 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 library: Here is how to use a pre-trained model to upscale your image: Training data The models for 2x, 3x and 4x image super resolution were pretrained on 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.. 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. We need the huggingface datasets library to download the data: The following code gets the data and preprocesses/augments the data. Pretraining The model was trained on GPU. The training code is provided below: Evaluation results The evaluation metrics include PSNR and SSIM. Evaluation datasets include: Set5 - Bevilacqua et al. (2012) Set14 - Zeyde et al. (2010) BSD100 - Martin et al. (2001) Urban100 - Huang et al. (2015) The results columns below are represented below as PSNR/SSIM. They are compared against a Bicubic baseline. |Dataset |Scale |Bicubic |pan | |--- |--- |--- |--- | |Set5 |2x |33.64/0.9292 |37.77/0.9599 | |Set5 |3x |30.39/0.8678 |34.64/0.9376 | |Set5 |4x |28.42/0.8101 |31.92/0.8915 | |Set14 |2x |30.22/0.8683 |33.42/0.9162 | |Set14 |3x |27.53/0.7737 |30.8/0.8544 | |Set14 |4x |25.99/0.7023 |28.57/0.7802 | |BSD100 |2x |29.55/0.8425 |33.6/0.9235 | |BSD100 |3x |27.20/0.7382 |29.47/0.815 | |BSD100 |4x |25.96/0.6672 |28.35/0.7595 | |Urban100 |2x |26.66/0.8408 |31.31/0.9197 | |Urban100 |3x | |28.61/0.8603 | |Urban100 |4x |23.14/0.6573 |25.63/0.7692 | You can find a notebook to easily run evaluation on pretrained models below: BibTeX entry and citation info

Read more

$-/run

484

Huggingface

pan-bam

pan-bam

Pixel Attention Network (PAN) PAN 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 Efficient Image Super-Resolution Using Pixel Attention by Zhao et al. (2020) and first released in this repository. 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 and model upscaling. Model description The PAN model proposes a a lightweight convolutional neural network for image super resolution. Pixel attention (PA) is similar to channel attention and spatial attention in formulation. PA however produces 3D attention maps instead of a 1D attention vector or a 2D map. This attention scheme introduces fewer additional parameters but generates better SR results. This model also applies the balanced attention (BAM) method invented by Wang et al. (2021) to further improve the results. The model is very lightweight with the model being just 260k to 270k parameters (~1mb). 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 library: Here is how to use a pre-trained model to upscale your image: Training data The models for 2x, 3x and 4x image super resolution were pretrained on 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.. 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. We need the huggingface datasets library to download the data: The following code gets the data and preprocesses/augments the data. Pretraining The model was trained on GPU. The training code is provided below: Evaluation results The evaluation metrics include PSNR and SSIM. Evaluation datasets include: Set5 - Bevilacqua et al. (2012) Set14 - Zeyde et al. (2010) BSD100 - Martin et al. (2001) Urban100 - Huang et al. (2015) The results columns below are represented below as PSNR/SSIM. They are compared against a Bicubic baseline. |Dataset |Scale |Bicubic |pan-bam | |--- |--- |--- |--- | |Set5 |2x |33.64/0.9292 |37.7/0.9596 | |Set5 |3x |30.39/0.8678 |34.62/0.9371 | |Set5 |4x |28.42/0.8101 |31.9/0.8911 | |Set14 |2x |30.22/0.8683 |33.4/0.9161 | |Set14 |3x |27.53/0.7737 |30.83/0.8545 | |Set14 |4x |25.99/0.7023 |28.54/0.7795 | |BSD100 |2x |29.55/0.8425 |33.6/0.9234 | |BSD100 |3x |27.20/0.7382 |29.47/0.8153 | |BSD100 |4x |25.96/0.6672 |28.32/0.7591 | |Urban100 |2x |26.66/0.8408 |31.35/0.92 | |Urban100 |3x | |28.64/0.861 | |Urban100 |4x |23.14/0.6573 |25.6/0.7691 | You can find a notebook to easily run evaluation on pretrained models below: BibTeX entry and citation info

Read more

$-/run

396

Huggingface

Similar creators