/Mangaka-Drawing-Style-Recommendation-System

Primary LanguagePython

Mangaka Drawing Style Identication

Important Things

  • SlicingWithFaceDetectionAndTextExtraction.py is the last updated file to extract the features.

    input : The mangas from 109Manga data set

    It produces :

      1- features.csv (List of features)

  2- mangaNames.csv (List of corresponding Manga names)

  • MachineLearningReduced.py is the last updated file for machine learning.

    input :

      1- features.csv (List of features)

  2- mangaNames.csv (List of corresponding Manga names)

    output :

      - predicted.csv 

  	C1 : Manga Name

  	C2 : The first predicted Manga

  	C3 : The Second predicted Manga

  	C4 : The Third predicted Manga

  • visualization.py is the last updated file for visualizing each feature.

    input :

      1- features.csv (List of features)

  2- mangaNames.csv (List of corresponding Manga names)

    output :

      - Plots, Change FeatureIndex (Value from 0 -> N) to plot each feature.

  • Colleration.py is the last updated file for finding colleration for each feature.

    input :

      1- features.csv (List of features)

  2- mangaNames.csv (List of corresponding Manga names)

    output :

      - Collaration.csv (List of colleration value per feature).

  - CollarationMargin80%.csv (All features that have 80% colleration or more).

Machine Learning - Final

1- Machine Learning Neural (Neural network solution)
2- Output : predictedFinal (The best we could get ^_^)

Taken Features

1-  Feature_Page_Average_Area_To_Rectangle_Ratio.
2-  Feature_Page_Average_Slope.
3-  Feature_Page_Average_Canny_Lines.
4-  Feature_SceneCanny_Average_Pixels.
5-  Feature_Tone_Average_Pixels.
6-  Feature_Outlines_To_Tones_Pixels_Ratio.
7-  Density per unit area deviation.
8-  Density per unit area.
9-  FaceCanny (Vectors Pixels) Pixels per unit area.
10- Feature_FaceTone_Pixels.
11- Ratio FaceCanny / Face Tone.
12- numberOfGradientColorsInFaces.
13- numberOfGradientColorsInScenes.