CUDI

Description

CUDI is for Custom Diaporama. This app is a display of the world, it will show you images with a particular aesthetic to keep your attention on the screen.

Installation

Never forget to update your system:
```
sudo apt update && sudo apt upgrade
```
Install Rust and Cargo if it's not already done

Install PostgreSQL:

linux:
sudo apt install postgresql postgresql-client

windows:
setx PQ_LIB_DIR "C:\Program Files\PostgreSQL\15\lib"
install postgresql from their website

set this env

echo DATABASE_URL=postgres://username:password@localhost/diesel_demo > .env

Create the database:

CREATE ROLE gmx;
CREATE DATABASE cudi_db;
CREATE USER gmx WITH PASSWORD '1234';
ALTER ROLE gmx SET default_transaction_isolation TO 'read committed';
GRANT ALL PRIVILEGES ON DATABASE cudi_db TO gmx;
\q

Run following commands:

cd media_handler
diesel migration run
cd ..
python3 toolbox.py -fc/--from-scratch

To do

Steps

Simple OpenGL engine in Rust:
- ~~Open a window with some images shown~~
- ~~Handle the high volume of local image to download per-second~~
- Handle the sound
- Handle video on screen
Options and customizations of the engine:
- Custom parameter to control the engine
  - ~~FPS~~
  - Filter
  - Mode (Cinema/Background/image with cudi in a plain rect see data/readme)
  - ~~Media path selector~~
  - Tags (WIP)
  - Animations
- Custom parameter in UX
- Show Music player
- Show Music analysis window
- Media history (like/dislike) -> search engine w/ postgresql (WIP)
External API (real-time data acquisition):
- Download photos or videos from API such as Archillect, Tumblr ou Google Photos
- Handle the Deezer or Youtube API to play music
ML everywhere:
- Add auto tagging image/video with themes/colors
- Add music tagging with themes/styles and BPM selector
- Adapt music and image/video BPM and tags

Main parts:

Image selector

create a database to stock the link of the image
create an ImageLink class that get the link from Archillect or Google Photos or instagram feed (check the max requests)
create an ML tag creator class using tensorflow to make a classification from all images
found different sources to get the image's links

ML Tag

Image tagger to reference images or videos following recurrent themes or aesthetic descriptions. The tags could be stored within a SQL database next to the image link.

Find tag that referenced this themes:

Main Colors
Feelings
Vibes / mood
Art mouvements
Words
Custom tag to create KNN like media groups
Music style

Screen handler

create a Menu class that override the screen and manage all the options (image_speed - tags - mode - filter - media_root)
create different mode (normal - gif surrounding the screen - animation (image or video where a part of the screen will aff cudi, the reste will be the template))
add an history of every file with a tag search or color search

Music handler

manage the audio from the computer
analyze the sound's curb to adapt cudi's parameter
add an ML class to adapt the tags from the audio

Animation

Animation could be add to the screen:

TV screen (high speed video* accelerated with b/w filter and speed cut)
Images in line on the top of the screen, a slow dark wallscreen animation will be always display. Those images will describe a type of sound (frequencies). When this type will appear in a music the corresponding image will grow then retake is place. All the sounds will create a nearly chaotic and unique animation for each music.
Color palette with gradient from left to right in high speed
Fill the screen with the same image in high speed and in a geometric way
Themes displayed with associative color palette (death -> dark)
pixel sorting
video in background and cudi on the top on certains part of the screen (anonymisation instead of blurring)
check compression artifact here also

xaviave/cudi_rs