Urban Scene Editing with Diffusion Model using Segmentation Mask

With the growing interest in autonomous driving systems, the demand for real-world driving environment datasets is on the rise. However, traditional data collection methods are time-consuming and costly, and struggle to capture a variety of situations. This paper addresses these issues by proposing a diffusion-based image-to-image translation model that utilizes image segmentation label information. The model effectively separates parts of the image for translation while preserving other information. We validate the effectiveness of the proposed model on Cityscapes dataset, a widely-used urban scene dataset.