Automatic Colorizing Black and White Image System Using Deep Learning

Colorization involves transforming a grayscale (black-and-white) image into a colorized version that reflects the semantic color tones of the original input. Over recent years, this process has garnered substantial attention, with significant advancements achieved by researchers in the field. The technique has numerous applications, such as in medical imaging and the restoration of historical artifacts. Various methods have been developed to address this challenge, leveraging technologies like Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs). These models differ in their architectural frameworks and are evaluated using diverse datasets. This paper explores some of these methodologies, comparing the performance of generative models with traditional deep neural networks, while also highlighting existing limitations. Additionally, it provides a concise overview of past and current progress in the domain of image colorization, drawing on contributions from numerous researchers and studies. a widely recognized deep learning algorithm. The proposed model processes grayscale images as input and predicts their corresponding colors based on training data. The Lab color space is employed in this approach, where the model receives the L channel (lightness) as input and predicts the ab channels (color information) as output. A subset of the ImageNet dataset was used for training and evaluation, consisting of 39,604 randomly selected images, divided into 80% for training and 20% for testing.