Talk at ECVP 2019

ECVP is the European Conference on Visual Perception. In 2019, it took place in Leuven, Belgium. This talk presented the preliminary results of my research on the orientation perception and judgement of abstract paintings.

Abstract

Artistic composition (the structural organization of pictorial elements) is often characterized by some basic rules and heuristics, but art history does not offer quantitative tools for e.g. segmenting individual elements and measuring their interactions. To discover whether a metric description of this kind is even possible, we exploit a deep-learning algorithm that attempts to capture the perceptual mechanism underlying composition in humans. We rely on a robust behavioral marker with known relevance to higher-level vision: orientation judgements i.e. telling whether a painting is hung "right-side up". Humans can perform this task even for abstract paintings. To account for this finding, existing models rely on arbitrarily selected recognizable content and/or specific image statistics, often in accordance with explicit rules from art theory. Our approach does not commit to any such assumptions/schemes, yet it outperforms previous models as well as human observers tested in a web-based experiment. It appears that our model captures more compositional regularities than the average human observer. Interestingly, the more abstract the painting, the more our model relies on deeper layers. We attempt to interpret this result via filter visualization techniques that expose the strategy adopted by the model.