Images of Image Machines. Visual Interpretability in Computer Vision for Art

摘要

Despite the emergence of interpretable machine learning as a distinct area of research, the role and possible uses of interpretability in digital art history are still unclear. Focusing on feature visualization as the most common technical manifestation of visual interpretability, we argue that in computer vision for art visual interpretability is desirable, if not indispensable. We propose that feature visualization images can be a useful tool if they are used in a non-traditional way that embraces their peculiar representational status. Moreover, we suggest that exactly because of this peculiar representational status, feature visualization images themselves deserve more attention from the computer vision and digital art history communities.