Artistic Domain Generalisation Methods are Limited by their Deep Representations
arXiv:1907.12622
The paper investigates why artistic domain generalisation methods struggle, showing that fixing AlexNet's final layer to random values matches state-of-the-art performance on the PACS benchmark, and argues that texture alone is insufficient for cross-depiction recognition, requiring higher-order shape and structure representations.
Abstract
The cross-depiction problem refers to the task of recognising visual objects regardless of their depictions; whether photographed, painted, sketched, {\em etc}. In the past, some researchers considered cross-depiction to be domain adaptation (DA). More recent work considers cross-depiction as domain generalisation (DG), in which algorithms extend recognition from one set of domains (such as photographs and coloured artwork) to another (such as sketches). We show that fixing the last layer of AlexNet to random values provides a performance comparable to state of the art DA and DG algorithms, when tested over the PACS benchmark. With support from background literature, our results lead us to conclude that texture alone is insufficient to support generalisation; rather, higher-order representations such as structure and shape are necessary.
7 pages, 3 figures, 1 Table, 3 Algorithms