Application and Analysis of Black and White Image Coloring Based on Generative Adversarial Networks (GANs)

Ming Him Foun

2024

Abstract

This paper demonstrates a comprehensive study on the application of Generative Adversarial Networks (GANs) for colorizing black-and-white images, employing the extensive COCO dataset for training and evaluating various deep learning frameworks. By integrating U-Net architecture with Residual Network (ResNet) 18 and Visual Geometry Group (VGG) 16 backbones within a PatchGAN framework, the study proposes a sophisticated method for adding color to grayscale images, aiming to create visually compelling and aesthetically pleasing results. The research adopts a systematic approach, beginning with image resizing and conversion to the Commission Internationale Eclairage lab (CIELAB) color space, followed by generator pretraining and subsequent PatchGAN training to finalize the colorization process. Through extensive experimentation, the study assesses the performance of the proposed models, revealing that the U-Net generator enhanced with a ResNet18 backbone significantly outperforms the VGG16 counterpart across multiple metrics, including Mean Squared Error (MSE), with a score of 1446.38961, Color Structural Similarity Index Measure (Color SSIM) of 0.87444, and 3.28116 for CIEDE2000. Despite building upon existing codes and frameworks, this study significantly advances the discourse in deep learning-based image colorization, emphasizing the comparative performance of different architectural choices and paving the way for future enhancements in the field.

Paper Citation

in Harvard Style

Him Foun M. (2024). Application and Analysis of Black and White Image Coloring Based on Generative Adversarial Networks (GANs). In Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI; ISBN 978-989-758-713-9, SciTePress, pages 354-361. DOI: 10.5220/0012938100004508

in Bibtex Style

@conference{emiti24,
author={Ming Him Foun},
title={Application and Analysis of Black and White Image Coloring Based on Generative Adversarial Networks (GANs)},
booktitle={Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI},
year={2024},
pages={354-361},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012938100004508},
isbn={978-989-758-713-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI
TI - Application and Analysis of Black and White Image Coloring Based on Generative Adversarial Networks (GANs)
SN - 978-989-758-713-9
AU - Him Foun M.
PY - 2024
SP - 354
EP - 361
DO - 10.5220/0012938100004508
PB - SciTePress