Fine-Grained Semantic Image Synthesis with Object-Attention Generative Adversarial Network

Semantic image synthesis is a new rising and challenging vision problem accompanied by the recent promising advances in generative adversarial networks. The existing semantic image synthesis methods only consider the global information provided by the semantic segmentation mask, such as class label, global layout, and location, so the generative models cannot capture the rich local fine-grained information of the images (e.g., object structure, contour, and texture). To address this issue, we adopt a multi-scale feature fusion algorithm to refine the generated images by learning the fine-grained information of the local objects. We propose OA-GAN, a novel object-attention generative adversarial network that allows attention-driven, multi-fusion refinement for fine-grained semantic image synthesis. Specifically, the proposed model first generates multi-scale global image features and local object features, respectively, then the local object features are fused into the global image features to improve the correlation between the local and the global. In the process of feature fusion, the global image features and the local object features are fused through the channel-spatial-wise fusion block to learn ‘what’ and ‘where’ to attend in the channel and spatial axes, respectively. The fused features are used to construct correlation filters to obtain feature response maps to determine the locations, contours, and textures of the objects. Extensive quantitative and qualitative experiments on COCO-Stuff, ADE20K and Cityscapes datasets demonstrate that our OA-GAN significantly outperforms the state-of-the-art methods.

An Observational Study of Social Interaction Skills and Behaviors in Cornelia de Lange, Fragile X and Rubinstein-Taybi Syndromes

We directly assessed the broader aspects of sociability (social enjoyment, social motivation, social interaction skills and social discomfort) in individuals with Cornelia de Lange (CdLS), fragile X (FXS) and Rubinstein-Taybi syndromes (RTS), and their association with autism characteristics and chronological age in these groups. Individuals with FXS (p < 0.01) and RTS (p < 0.01) showed poorer quality of eye contact compared to individuals with CdLS. Individuals with FXS showed less person and more object attention than individuals with CdLS (p < 0.01). Associations between sociability and autism characteristics and chronological age differed between groups, which may indicate divergence in the development and aetiology of different components of sociability across these groups. Findings indicate that individuals with CdLS, FXS and RTS show unique profiles of sociability.

17. Within-Object Attention: Attentional Concentration and Amplification in Moving Versus Static Conditions

When presented with multiple stimuli, attention serves to help us select which items to focus on for further processing. The techniques we apply to determine where to focus within an object are less clearly understood than those applied for selecting objects. This study aims to observe if there are any existing trends in how we allocate our attention within objects, as well as to examine how different factors affect this cognitive process. Two previously identified trends in how attention is allocated within lines are attentional concentration – a tendency to focus on the center of the lines – and attentional amplification – the increase in center focus as line length increases (Alvarez and Scholl, 2005). Object dynamics is a variable that can be manipulated to test if these effects represent a higher-order tracking strategy, or whether they will still be exhibited when presented with stationary lines. This study examines whether these effects can be replicated when tested with moving lines as well as stationary lines. 22 participants completed tracking tasks, either moving or static, where they tracked two designated target lines while simultaneously watching for the appearance of a probe at either the center or end of any of the lines. In the moving condition, probe detection was significantly lower at the ends of lines (indicating attentional concentration), and this decrease in performance grew slightly as line length increased (indicating weak attentional amplification); however, no significant differences in probe detection between centers and ends of lines were found in the static condition.