Memoria

Memoria is a tool to help you sort, categorize, and generally assist with making sense of scanned images. It focusses on dealing with scanned, pre-digital images, such as slides and photos, which are lacking the detailed EXIF metadata (like precise GPS coordinates or exact capture dates) that many modern photo library applications rely upon.

Memoria provides the tools to manually enrich this metadata, allowing you to catalogue, rediscover, and share your memories with the context they deserve.

Naming

Memoria can be translated from Latin as "memory", as that is what the project aims to help preserve and share with your family.

Inspiration

This project was born out of another project I embarked on in late 2023, the scanning of thousands of slides from my parents' and grandparents' trips and life. As I accumulated more and more images, with pictures of my family, their trips, and general life, I found it hard to keep track of the images of people in particular. Then I wanted to know where the picture was taken, roughly what it was about, an idea of the year, and maybe even month.

But I wasn't around for most of the images, so I had no idea about most of that. And so, this project was born with the goal to assist one or more family members in sorting out all that information.

It allows one or more users (probably other family members) to view, sort, categorize old photographs, slides, etc and then share with more users to view, remember and enjoy a little bit of history no one might have ever seen before. Because, after all, who has a slide projector anymore?

Features

• Metadata Working Group Support
  • Face & Pet Support
    • Uses the MWG Face & Pet regions, ties people and pets to the images where they are found
    • Includes the bounding box, which is displayed in the frontend
  • Rough Location:
    • Associate images with geographical locations—from broad country data down to specific sub-locations. This is ideal for images where only approximate places are known.
    • Uses the MWG location tags to build a rough idea of location for the image
    • If the MWG locations are not set, falls back to a tag based attempt to gather location information
  • Rough Date
    • Assign approximate or exact dates (year, month, day) with validity flags, perfect for when only a general timeframe is recalled.
    • Uses a tag based structure to store a rough idea of when the image was originally captured.
  • Description: Allows users to add image descriptions, stored per the MWG
  • Title: Although the MWG doesn't specify a title field, it is possible to set an image's title as well, delegating to EXIFTool what that sets
• Albums:
  • Group your images into thematic albums.
  • Crucially, manually order images within albums. This feature is perfect for reconstructing the original sequence of digitized slide carousels or the narrative flow of physical photo albums.
• Multi-User & Multi-Family Support:
  • Securely manage access for multiple individual users or distinct family groups.
  • Utilizes a robust group-based permissions system for images, albums, and other metadata, ensuring that private collections remain private and accessible only to authorized individuals or groups.
• Robust Image Handling:
  • Efficiently indexes your image library.
  • Generates and serves optimized image versions (thumbnails, larger web-display copies) for fast and responsive viewing, while preserving your originals.
  • Ensures metadata consistency by synchronizing database information with physical image files where appropriate, adhering to MWG standards.
• Modern & Intuitive Interface:
  • Experience a clean, responsive user interface built with React Bootstrap.
  • Choose between light and dark themes, with an option to follow your system's preference.
  • Navigate and manage your photos with ease

Bugs

• The location information is loaded in the image detail, even if the location editing modal is not open

Roadmap

Markdown Support

• Make descriptions treated as Markdown text, with rendering and editor support
• Allows more robust descriptions, links, etc

Bulk Operation Support

• Bulk editing of permissions
• Bulk adding to albums
• Album sorting of sections of selected images

More Region Support

• Add, edit and delete boxes for faces and pets in images

Frontend Error Handling

• Standardize error handling, using toasts or similar
• Break out the problem fields on forms to display more helpful text

Advanced Searching

• Text based searching over description and captions
• Filter on images containing any people or all people
• Filter for images containing Person A & Person B IN Place between YEAR & YEAR
• Consider Mellisearch or similar tools

AI Integration

• Semantic searching over descriptions and titles
• Automatic face tagging and recognition
  • Using either local or cloud AI models to find and recognize faces

Task Status

• Integrate with Huey signals to provide task status and tracking
• Consider integrating websockets for more real-time status

Tasks from Frontend

• Allow users to start an index, sync, etc task manually via the frontend

Next & Previous

• Allow walking through all images in an album or folder using next and previous
• Potentially extend to include paginated all images display

Documentation

• Add documentation for getting started
• Document the capabilities of each type of user, group, etc

Async Support

• Given Django's current async support, look for opportunities to utilize it, either in the current state or future releases

Testing

• Currently have PoC for testing the backend, needs to be expanded
• Determine how to test the frontend
• Integrate code coverage, including branches

Path Guessing Protection

Currently, thumbnails and the larger, but compressed images are stored based on their integer primary key. Although the default nginx configuration prevents walking that path, an outside user could easily guess the paths, given a few samples. If the application is not protected by something like Traefik + Authelia, this could allow an unauthroized user to access the images.

Possible solutions:

Path signing

1. Frontend needs an image.
2. Frontend makes an authenticated request to your Django backend for the image URL.
3. Django generates a signed URL (e.g., /media/image.jpg?signature=XYZ&expires=123). This signature is a hash of the URL and a secret key, plus an expiration timestamp.
4. Django returns this signed URL to the frontend.
5. Frontend uses this signed URL to request the image from NGINX.
6. NGINX receives the request, parses the signature and expiration, and verifies them using its own copy of the secret key.
7. If the signature is valid and not expired, NGINX serves the file. Otherwise, it denies access.

Useful: TimestampSigner?

Cons:

• Does need to share a secret key with nginx?

Internal Nginx Path

1. Frontend requests /media/image.jpg.
2. NGINX receives the request.
3. Instead of directly serving the file, NGINX passes the request to your Django backend (e.g., /api/media/image.jpg).
4. Django authenticates the user, checks permissions for image.jpg.
5. If authorized, Django sends an internal redirect to NGINX (more efficient for large files) using X-Accel-Redirect or X-Sendfile headers.

Tech Stack

• Backend:
  • Python
  • Django & Django Ninja (for REST APIs)
  • PostgreSQL (Database)
  • Valkey (Redis fork, for caching/Huey broker)
  • Huey (Task Queue)
• Frontend:
  • TypeScript
  • React
  • Vite (Build Tool)
  • pnpm (Package Manager)
  • React Bootstrap & Bootstrap Icons (UI)
  • Axios (HTTP Client)
• DevOps & Tooling:
  • Docker & Docker Compose
  • MkDocs (Documentation)
  • Go Task (Task Runner)