Image Colorizer

In this project, the aim was to tackle the challenging problem of image colorization, where the goal is to predict the color at each pixel given a grayscale image. The dataset utilized for this endeavor was the CIFAR-10 dataset, which consists of 32x32 pixel images. The task involved transforming these images from color to grayscale and then attempting to recolor them. Initially, a convolutional autoencoder was employed to address this problem, leveraging its ability to learn efficient data codings in an unsupervised manner. This approach was further refined by experimenting with various modifications, including the introduction of skip connections to enhance the model's performance. Later in the project, the focus shifted to exploring Conditional Generative Adversarial Networks (cGANs), comparing their effectiveness against autoencoders and investigating additional techniques to optimize the training process. Through these exercises, the project aimed to deepen understanding of advanced neural network architectures and their application in image processing tasks.