Neurons in the pigeon visual network discriminate between faces, scrambled faces, and sine grating images

Discriminating between object categories (e.g., conspecifics, food, potential predators) is a critical function of the primate and bird visual systems. We examined whether a similar hierarchical organization in the ventral stream that operates for processing faces in monkeys also exists in the avian visual system. We performed electrophysiological recordings from the pigeon Wulst of the thalamofugal pathway, in addition to the entopallium (ENTO) and mesopallium ventrolaterale (MVL) of the tectofugal pathway, while pigeons viewed images of faces, scrambled controls, and sine gratings. A greater proportion of MVL neurons fired to the stimuli, and linear discriminant analysis revealed that the population response of MVL neurons distinguished between the stimuli with greater capacity than ENTO and Wulst neurons. While MVL neurons displayed the greatest response selectivity, in contrast to the primate system no neurons were strongly face-selective and some responded best to the scrambled images. These findings suggest that MVL is primarily involved in processing the local features of images, much like the early visual cortex.

Body shape as a visual feature: evidence from spatially-global attentional modulation in human visual cortex

Feature-based attention supports the selection of goal-relevant stimuli by enhancing the visual processing of attended features. A defining property of feature-based attention is that it modulates visual processing beyond the focus of spatial attention. Previous work has reported such spatially-global effects for low-level features such as color and orientation, as well as for faces. Here, using fMRI, we provide evidence for spatially-global attentional modulation for human bodies. Participants were cued to search for one of six object categories in two vertically-aligned images. Two additional, horizontally-aligned, images were simultaneously presented but were never task-relevant across three experimental sessions. Analyses time-locked to the objects presented in these task-irrelevant images revealed that responses evoked by body silhouettes were modulated by the participants' top-down attentional set, becoming more body-selective when participants searched for bodies in the task-relevant images. These effects were observed both in univariate analyses of the body-selective cortex and in multivariate analyses of the object-selective visual cortex. Additional analyses showed that this modulation reflected response gain rather than a bias induced by the cues, and that it reflected enhancement of body responses rather than suppression of non-body responses. Finally, the features of early layers of a convolutional neural network trained for object recognition could not be used to accurately categorize body silhouettes, indicating that the fMRI results were unlikely to reflect selection based on low-level features. These findings provide the first evidence for spatially-global feature-based attention for human bodies, linking this modulation to body representations in high-level visual cortex.

Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild (Extended Abstract)

We propose a method to learn 3D deformable object categories from raw single-view images, without external supervision. The method is based on an autoencoder that factors each input image into depth, albedo, viewpoint and illumination. In order to disentangle these components without supervision, we use the fact that many object categories have, at least approximately, a symmetric structure. We show that reasoning about illumination allows us to exploit the underlying object symmetry even if the appearance is not symmetric due to shading. Furthermore, we model objects that are probably, but not certainly, symmetric by predicting a symmetry probability map, learned end-to-end with the other components of the model. Our experiments show that this method can recover very accurately the 3D shape of human faces, cat faces and cars from single-view images, without any supervision or a prior shape model. Code and demo available at https://github.com/elliottwu/unsup3d.

One-Shot Affordance Detection

Affordance detection refers to identifying the potential action possibilities of objects in an image, which is an important ability for robot perception and manipulation. To empower robots with this ability in unseen scenarios, we consider the challenging one-shot affordance detection problem in this paper, i.e., given a support image that depicts the action purpose, all objects in a scene with the common affordance should be detected. To this end, we devise a One-Shot Affordance Detection (OS-AD) network that firstly estimates the purpose and then transfers it to help detect the common affordance from all candidate images. Through collaboration learning, OS-AD can capture the common characteristics between objects having the same underlying affordance and learn a good adaptation capability for perceiving unseen affordances. Besides, we build a Purpose-driven Affordance Dataset (PAD) by collecting and labeling 4k images from 31 affordance and 72 object categories. Experimental results demonstrate the superiority of our model over previous representative ones in terms of both objective metrics and visual quality. The benchmark suite is at ProjectPage.

Multiclass semantic segmentation for digitisation of movable heritage using deep learning techniques

<p class="VARAbstract">Digitisation processes of movable heritage are becoming increasingly popular to document the artworks stored in our museums. A growing number of strategies for the three-dimensional (3D) acquisition and modelling of these invaluable assets have been developed in the last few years. Their objective is to efficiently respond to this documentation need and contribute to deepening the knowledge of the masterpieces investigated constantly by researchers operating in many fieldworks. Nowadays, one of the most effective solutions is represented by the development of image-based techniques, usually connected to a Structure-from-Motion (SfM) photogrammetric approach. However, while images acquisition is relatively rapid, the processes connected to data processing are very time-consuming and require the operator’s substantial manual involvement. Developing deep learning-based strategies can be an effective solution to enhance the automatism level. In this research, which has been carried out in the framework of the digitisation of a wooden maquettes collection stored in the ‘Museo Egizio di Torino’, using a photogrammetric approach, an automatic masking strategy using deep learning techniques is proposed, to increase the level of automatism and therefore, optimise the photogrammetric pipeline. Starting from a manually annotated dataset, a neural network was trained to automatically perform a semantic classification to isolate the maquettes from the background. The proposed methodology allowed the researchers to obtain automatically segmented masks with a high degree of accuracy. The workflow is described (as regards acquisition strategies, dataset processing, and neural network training). In addition, the accuracy of the results is evaluated and discussed. Finally, the researchers proposed the possibility of performing a multiclass segmentation on the digital images to recognise different object categories in the images, as well as to define a semantic hierarchy to perform automatic classification of different elements in the acquired images.</p><p><strong>Highlights:</strong></p><ul><li><p>In the framework of movable heritage digitisation processes, many procedures are very time-consuming, and they still require the operator’s substantial manual involvement.</p></li><li><p>This research proposes using deep learning techniques to enhance the automatism level in the generation of exclusion masks, improving the optimisation of the photogrammetric procedures.</p></li><li><p>Following this strategy, the possibility of performing a multiclass semantic segmentation (on the 2D images and, consequently, on the 3D point cloud) is also discussed, considering the accuracy of the obtainable results.</p></li></ul>

Red Collections in Contemporary China

“Red Collecting” is a widespread phenomenon in contemporary China. It refers to the collecting of objects from the Chinese Communist Party’s history. Red Collecting has received only minimal treatment in English-language scholarly literature, much of which focuses on individual object categories (primarily propaganda posters and Chairman Mao badges) and overemphasises the importance of Cultural Revolution objects within the field. Because of this limited focus, the collectors’ motivations have been similarly circumscribed, described primarily in terms of either neo-Maoist nostalgia or the pursuit of profit. This article will seek to enhance this existing literature and, in doing so, offer a series of new directions for research. It makes two main arguments. First, that the breadth of objects incorporated within the field of Red Collecting is far broader than current literature has acknowledged. In particular, the importance of revolutionary-era (pre-1949) collections, as well as regional and rural collections is highlighted. Second, it argues that collectors are driven by a much broader range of motivations, including a variety of both individual and social motivations. Significantly, it is argued that collectors’ intentions and their understandings of the past do not always align; rather, very different understandings of China’s recent past find expression through Red Collecting. As such, it is suggested that Red Collecting constitutes an important part of contemporary China’s “red legacies,” one which highlights the diversity of memories and narratives of both the Mao era and the revolutionary period. Image © Hou Feng