Term Project of CS 485 Deep Generative Networks Course
Image Colorization using U-Net with Skip Connections is an advanced deep learning project that focuses on generating plausible color versions of grayscale images. This repository contains the implementation of a U-Net-based generator with skip connections, trained using Generative Adversarial Networks (GANs), to achieve accurate and visually appealing colorization results.
- U-Net-based generator with skip connections for high-quality colorization.
- Integration of attention modules to preserve semantic information during colorization.
- Three methods for achieving diverse colorizations based on style transfer principles.
- Evaluation metrics such as FID, SSIM, and color histogram KL-Divergence.