Learning Portrait Style Representations
This work addresses the problem of computationally understanding high-level artistic style for art historians and computer vision researchers, offering an incremental step towards more nuanced art analysis.
This paper explores learning high-level style representations in portrait artwork using neural networks. It investigates the impact of art historian-annotated triplets, statistical priors, and photo realism priors on learned style features, finding that these align with art historical research and enable zero-shot artist classification.
Style analysis of artwork in computer vision predominantly focuses on achieving results in target image generation through optimizing understanding of low level style characteristics such as brush strokes. However, fundamentally different techniques are required to computationally understand and control qualities of art which incorporate higher level style characteristics. We study style representations learned by neural network architectures incorporating these higher level characteristics. We find variation in learned style features from incorporating triplets annotated by art historians as supervision for style similarity. Networks leveraging statistical priors or pretrained on photo collections such as ImageNet can also derive useful visual representations of artwork. We align the impact of these expert human knowledge, statistical, and photo realism priors on style representations with art historical research and use these representations to perform zero-shot classification of artists. To facilitate this work, we also present the first large-scale dataset of portraits prepared for computational analysis.