Abstract:
Previous approaches to the colorization of grayscale images rely on human manual labor and often produce desaturated results that are not likely to be true colorizations....Show MoreMetadata
Abstract:
Previous approaches to the colorization of grayscale images rely on human manual labor and often produce desaturated results that are not likely to be true colorizations. Inspired by [15], we proposed an automatic approach based on deep neural networks to color the image in grayscale. We have studied several models, approaches and loss function to understand the best practices to produce a plausible colorization. We trained a convolutional neural network by noting that some loss functions work better than others. We used the VGG-16 CNN model based on the classification with the loss of cross entropy that is very well to produce a plausible colorization.
Published in: 2019 12th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)
Date of Conference: 19-21 October 2019
Date Added to IEEE Xplore: 23 January 2020
ISBN Information: