Scientific report on image augmentation with Generative AI

This report highlights the importance of synthetic data in machine learning for addressing challenges in data collection and annotation. The Confiance.ai program explored GAN-based and diffusion models to enhance datasets for computer vision tasks like object detection. Key approaches include image editing, image-to-image translation, and layout-to-image models. These methods were applied to datasets such as BDD100k, Mapillary, and rarePlanes, improving model robustness and data diversity.