An open source gallery web application with AI features executed on the browser

Features

• Multimedia management: Upload, organize and share albums to other users.
• Object detection: Search through your gallery by objects in the image by using YOLOv8.
• Image upscaling: Enhance low resolution images with Remacri.
• Image generation: Generate any image you want quickly with the power of Stable Diffussion Turbo.

Prerequisites

Client side

To execute most of the models contained in this project, you need to be running the following setup:

• Browser: Any browser that supports the WebGPU API. It can be a recent version of a Chromium-based browser (Chrome, Edge, Brave, etc.) or a nightly version of Firefox with the about:config flag dom.webgpu.enabled set to true. Test it here or inside the app. shader-f16 should be listed as a feature to be able to use image generation.

Linux users: You may need to enable the #enable-unsafe-webgpu flag in chrome://flags to use WebGPU.

• GPU: A dedicated/integrated GPU with at least enough VRAM/system RAM:
  • Image upscaling: Currently up to 2GiB for a max resolution of 362x362.
  • Image generation (Stable Diffusion Turbo): 4GiB for resolutions <= 256x256, 8GiB for resolutions <= 512x512, 16GB for resolutions above.

Server side

• Linux environment: The application was developed and tested in a Linux environment. It may work in other environments, but it is not guaranteed.
• Docker Compose: To run the database, application and web server in containerized environments (or just database in development profile).

For local development, you will also need:

• Python >=3.11: To run Django backend. Preferably in a virtual environment.
• NPM: To run frontend.

Installation

Docker

1. Clone the repository: git clone https://github.com/luismiaresse/galeria
2. Run scripts depending on desired profile:

• devserver.sh for development:
  • Install Python dependencies: cd app/backend && pip install -r requirements.txt
  • Install NPM dependencies: cd app/frontend && npm install
  • Run the script with chmod +x ./devserver.sh && ./devserver.sh and access the app at http://localhost:5173.
• prodserver.sh for production: Just run chmod +x ./prodserver.sh && ./prodserver.sh and go to https://www.galeria.software (or your custom domain). Note that you will need to provide SSL certificates and a SSL key in the server/ssl folder.

License

Licensed under AGPLv3. This should be in any fork or redistribution, and should credit contributors as well. See LICENSE for more information.