Abstract
People often create art by following an artistic workflow involving multiplestages that inform the overall design. If an artist wishes to modify an earlierdecision, significant work may be required to propagate this new decisionforward to the final artwork. Motivated by the above observations, we propose agenerative model that follows a given artistic workflow, enabling bothmulti-stage image generation as well as multi-stage image editing of anexisting piece of art. Furthermore, for the editing scenario, we introduce anoptimization process along with learning-based regularization to ensure theedited image produced by the model closely aligns with the originally providedimage. Qualitative and quantitative results on three different artisticdatasets demonstrate the effectiveness of the proposed framework on both imagegeneration and editing tasks.