PROJECT CODENAME ARTIO

What is Artio ?

Artio is a project aiming to create an app acting like a Shazam for paintings. You go in a museum or gallerie, you pull out our smartphone and launch Artio app, you scan a painting with the smartphone camera and the app gives you informations about the painting (Title, Author, Date, Explainations...)

Why is named Artio ?

Artio stands for ART REAL TIME IDENTIFICATION OPERATION
Artio is also for Art / IO (Input-Output), an analogic-digital encounter
Artio is also a celtic bear goddess so that's cool! That's why I also think the logo should be a bear

How does it work ?

The project is divided into parts :

These modules are organized as following :

           +---------------------+
           |      Android App    |
           |                     |
           | Painting Extraction |
           |                     |
           +---+-----+-----------+
               ^     |
               |     |
Server response|     |   API REQUEST
->JSON with    |     |    -> Extracted painting
usefull info   |     |
               |     |
               |     v
            +--+-----+------------+
            |   Server            |
            |                     |
            | Painting Matching   +----> +------------+
            |                     <------+            |
            +---------------------+      |  Database  |
                                         |            |
                                         +------------+

Painting Extraction Module

This module will be written in C++ and will use the openCV library.

For now, the organisation of the algorithm is as following :

Input -> Resize -> To Grey Scale -> Low-Pass Filtering -> Edge Detection -> Shape Detection -> Filter Rectangles -> Parralax Correction -> Cropping -> Brightness, Contrast, Color Correction -> Output

Not so bad, but to be improved.

TODO

Ressources

[http://artsy.github.io/blog/2014/09/24/using-pattern-recognition-to-automatically-crop-framed-art/] [http://www.pyimagesearch.com/2014/04/21/building-pokedex-python-finding-game-boy-screen-step-4-6/]

Painting Matching Module

HARD PART I don't really know how to do it efficiently. Either a statistical approach with comparison metrics like Normalized Cross Corellation or a neural network approach with Convolutional Neural Networks. Or something else, if you have ideas edit this file

TODO

Ressources

[https://ip.cadence.com/uploads/901/cnn_wp-pdf]

Android App

For now the app should just be a wrapper for the Painting Extraction Module (PEM). Later improvements with user accounts and stuff could be made

Server

The server should handle requests from the app and wrap the Painting Matching Module (PMM) To be discussed

Don't hesitate to edit this file with Your ideas, Todo Lists, References...