Skip to content

README.md

Latest commit

 

History

History
84 lines (56 loc) · 2.69 KB

README.md

File metadata and controls

84 lines (56 loc) · 2.69 KB

Auto Colorization of Gray Scale Image using CNN

Overview

This project implements automatic colorization of grayscale images using Convolutional Neural Networks (CNN). The repository includes both basic auto-colorization and an advanced Ethnicity Aware Autocolorization system that considers ethnic characteristics for more accurate and culturally sensitive colorization.

Complete Walkaround is covered in this blog: Auto-Colorization of Grayscale Images using CNN

Tools Used

  • Python 3
  • Keras
  • Numpy
  • Tensorflow

Features

  • Standard Auto-Colorization: Basic CNN-based colorization for general grayscale images
  • Ethnicity Aware Auto-Colorization: Advanced pipeline that detects and considers ethnic characteristics for improved colorization accuracy
  • Pre-trained models for immediate use
  • Comprehensive testing and evaluation

Repository Structure

  • Dataset: Training and testing datasets
  • Screenshots: Result demonstrations
  • result: Output colorized images
  • Ethnicity Aware Autocolorization: Advanced ethnic-aware implementation with specialized notebooks and models
  • Auto_color.ipynb: Main colorization notebook
  • model.h5: Pre-trained model
  • model.json: Model architecture

Dataset

Dataset is included in the folder named Dataset.

Ethnicity Aware Autocolorization

The Ethnicity Aware Autocolorization folder contains an advanced implementation that:

  • Detects ethnic characteristics in facial images
  • Applies culturally appropriate colorization
  • Includes specialized models and testing notebooks
  • Features comprehensive evaluation using LPIPS metrics
  • Contains multiple test datasets for different ethnic groups

Key files in this folder:

  • Colorization Final.ipynb: Main colorization pipeline
  • Ethnic Detection Final.ipynb: Ethnicity detection system
  • Final Testing Both Pipeline.ipynb: Combined testing pipeline
  • Colorize.h5 and ColorizeTuned.h5: Pre-trained models

Screenshot of Result

Left Column includes the input images and the right column includes the automatically colorized images using CNN.

Result 1

Result 2

Result 3

Getting Started

  1. Clone the repository
  2. Install required dependencies: pip install tensorflow keras numpy
  3. Run Auto_color.ipynb for basic colorization
  4. Explore the Ethnicity Aware Autocolorization folder for advanced features

Usage

For basic colorization:

jupyter notebook Auto_color.ipynb

For ethnicity-aware colorization:

cd "Ethnicity Aware Autocolorization"
jupyter notebook "Final Testing Both Pipeline.ipynb"