Interactive Colorization

Interactive colorization aims to reconstruct the color information of grayscale images, guided by user input such as color strokes, text descriptions, or example colors. Recent research focuses on improving the accuracy and efficiency of this process, employing advanced architectures like diffusion models and vision transformers to better propagate user hints and handle color inconsistencies. These advancements are improving the realism and controllability of colorization, with applications ranging from image editing to assisting in tasks like 3D object detection in autonomous driving by providing semantic cues from colorized LiDAR data.