DOCSEARCHbyAlgolia
Submit Docs

UpscalerJS

website·Indexed 32 days ago

https://upscalerjs.com

A JavaScript library for image upscaling and enhancement using TensorFlow.js. It supports super-resolution via models like ESRGAN (Thick, Medium, Slim, Legacy) and the MAXIM family for denoising, deraining, and low-light enhancement. The library provides implementations for the browser, Node.js, and Cloudflare Workers, with features including Web Worker integration, tensor manipulation, and model warming.

Tokens
18.3K
Snippets
118
Records
168
Agent score
53%

What's inside UpscalerJS

  1. Overview of MAXIM Dehazing Outdoor model

    MAXIM Dehazing Outdoor is a TensorFlow.js model designed for dehazing outdoor images. It is based on the MAXIM (Multi-Axis MLP for Image Processing) architecture, which uses a UNet-shaped hierarchical structure and spatially-gated MLPs to provide an efficient and flexible backbone for image processing tasks. The weights for this model were trained on the RESIDE dataset.

  2. Overview of MAXIM Deraining model

    MAXIM Deraining is a TensorFlow.js model used for removing rain from images. It is based on the MAXIM (Multi-Axis MLP for Image Processing) architecture, which uses a UNet-shaped hierarchical structure and spatially-gated MLPs to provide an efficient, general-purpose vision backbone for image processing. The weights for this model were trained on the Rain13K dataset.

  3. Overview of UpscalerJS for browser-based image upscaling

    UpscalerJS is a JavaScript library used for performing image upscaling directly in the browser. It is designed to be model-agnostic, allowing developers to use any trained model that can be converted to Tensorflow.js. The tool can reduce image sizes by up to 1/16th.
    # UpscalerJS is available on GitHub
https://github.com/thekevinscott/UpscalerJS

  4. Overview of MAXIM Deblurring model

    MAXIM Deblurring is a TensorFlow.js model designed for image deblurring. It is based on the MAXIM (Multi-Axis MLP for Image Processing) architecture, which uses a UNet-shaped hierarchical structure and spatially-gated MLPs to handle long-range interactions efficiently. The model weights were trained on the GoPro dataset.

  5. Overview of UpscalerJS for browser-based image upscaling

    UpscalerJS is a JavaScript tool used for image upscaling directly in the browser. It is designed to be model-agnostic, allowing developers to use any trained model that can be converted to TensorFlow.js. It can help reduce image sizes by up to 1/16th.
    # Source code available at
https://github.com/thekevinscott/UpscalerJS

  6. Overview of UpscalerJS for browser-based image upscaling

    UpscalerJS is a JavaScript tool used for image upscaling directly in the browser. It is designed to be model-agnostic, allowing developers to use any trained model that can be converted to Tensorflow.js. It can reduce image sizes by up to 1/16th.
    # GitHub repository for UpscalerJS
https://github.com/thekevinscott/UpscalerJS

  7. Overview of MAXIM Denoising model

    MAXIM Denoising is a Tensorflow.js model used for denoising images. It is based on the MAXIM (Multi-Axis MLP for Image Processing) architecture, which uses a UNet-shaped hierarchical structure and spatially-gated MLPs to provide an efficient, general-purpose vision backbone for image processing tasks. The weights for this model were trained on the SIDD dataset.

  8. Overview of UpscalerJS

    UpscalerJS is an npm package designed to be agnostic to the specific upscaling model used. This allows for the swapping or improvement of models and the introduction of models tuned for specific use cases, such as faces or illustrations. Models are typically served via JS CDNs.

  9. Overview of UpscalerJS for browser-based image upscaling

    UpscalerJS is a JavaScript library used for performing image upscaling directly in the browser. It allows developers to reduce image sizes by up to 1/16th and is designed to be model-agnostic, meaning any trained model that can be converted to Tensorflow.js can be used with the tool.
    # UpscalerJS is available on GitHub
https://github.com/thekevinscott/UpscalerJS

  10. Overview of UpscalerJS

    UpscalerJS is an open-source JavaScript tool used for enhancing images using AI, specifically focusing on super-resolution (increasing image size while maintaining quality by using Neural Networks to 'paint in' expanded sections). It is built on top of TensorFlow.js and ships with built-in models, eliminating the need for separate model installations. It is compatible with the following environments:

    • Web Browsers
    • Node.js
    • Electron
    • Cloudflare Workers

  11. Overview of ESRGAN Thick models

    ESRGAN Thick models are high-quality upscaling models based on the Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) architecture. They provide superior visual quality and realistic textures compared to standard SRGAN, though they have slower inference speeds. They are best suited for Node.js environments with a GPU, but can be used in the browser with significant latency.

  12. Overview of UpscalerJS for browser-based image upscaling

    UpscalerJS is a model-agnostic JavaScript library that enables image super-resolution (upscaling) directly in the browser. It allows developers to use any trained model that can be converted to TensorFlow.js. By performing inference on the client side, developers can serve significantly smaller images (e.g., reducing file size by up to 1/16th) and reconstruct them to a higher resolution without the blurriness associated with bicubic interpolation.