In many critical investigations of machine vision, the focus lies almost exclusively on dataset bias and on fixing datasets by introducing more and more diverse sets of images. We propose that machine vision systems are inherently biased not only because they rely on biased datasets but also because their perceptual topology, their specific way of representing the visual world, gives rise to a new class of bias that we call perceptual bias. Concretely, we define perceptual topology as the set of those inductive biases in machine vision systems that determine its capability to represent the visual world. Perceptual bias, then, describes the difference between the assumed “ways of seeing” of a machine vision system, our reasonable expectations regarding its way of representing the visual world, and its actual perceptual topology. We show how perceptual bias affects the interpretability of machine vision systems in particular, by means of a close reading of a visualization technique called “feature visualization”. We conclude that dataset bias and perceptual bias both need to be considered in the critical analysis of machine vision systems and propose to understand critical machine vision as an important transdisciplinary challenge, situated at the interface of computer science and visual studies/Bildwissenschaft.

Digital Cultural Heritage and Digital Humanities are, historically seen, in focus of different communities as well as approaching different research topics and - from an organizational point of view - departments. However, are they that different? The idea of this joint article involving digital humanists and heritage researchers is to examine communities, concepts and research applications as well as shared challenges. Beyond a collection of problem-centred essays this is intended to initiate a fruitful discussion about commonalities and differences between both scholarly fields as well as to assess to which extent they are two sides of the same medal.

Human pose estimation (HPE) is a central part of understanding the visual narration and body movements of characters depicted in artwork collections, such as Greek vase paintings. Unfortunately, existing HPE methods do not generalise well across domains resulting in poorly recognised poses. Therefore, we propose a two step approach: (1) adapting a dataset of natural images of known person and pose annotations to the style of Greek vase paintings by means of image style-transfer. We introduce a perceptually-grounded style transfer training to enforce perceptual consistency. Then, we fine-tune the base model with this newly created dataset. We show that using style-transfer learning significantly improves the SOTA performance on unlabelled data by more than 6% mean average precision (mAP) as well as mean average recall (mAR) . (2) To improve the already strong results further, we created a small dataset (ClassArch) consisting of ancient Greek vase paintings from the 6–5th century BCE with person and pose annotations. We show that fine-tuning on this data with a style-transferred model improves the performance further. In a thorough ablation study, we give a targeted analysis of the influence of style intensities, revealing that the model learns generic domain styles. Additionally, we provide a pose-based image retrieval to demonstrate the effectiveness of our method. The code and pretrained models can be found at https://github.com/angelvillar96/STLPose .

The last few years have seen an explosion of medieval images in digital form, chiefly as a result of photo-library and manuscript digitisation projects. An entire corpus of images, even selected solely by scene or iconography, becomes an unwieldy object of study by traditional art-historical means. This is even more the case for medieval images, where authorship and dating are often cloudy and unclear, and the image itself is in many cases the first resource for scholarly inquiry.We take the digital image – in particular, the digital image of the body – as our object of study in a wide-ranging computationally-augmented reading of an image-corpus; ours is made up of thousands of depictions of the ‘Annunciation’ and ‘Baptism’, selected not only for their primacy in Christian art but for their dialogical interaction. Our corpus of 6,564 ‘Annunciations’ and 883 ‘Baptisms’, whilst not necessarily representative in density, includes a wide range of stylistic, theological and historical tendencies.We computationally extract not only body images but poses, gestures and interactions. Such a range of gestures allows for a morphological treatment of bodily motifs, whose multi-dimensional, quantitative nature allows us to complicate and problematise iconographic taxonomies, populating the spaces between categories. Finally, our gestural manifolds provide a morphological pointer to dissecting the microtemporalities of the scenes, and their relative dynamics and inconsistencies.