Artwork Style Recognition Using Vision Transformers and MLP Mixer

Through the extensive study of transformers, attention mechanisms have emerged as potentially more powerful than sequential recurrent processing and convolution. In this realm, Vision Transformers have gained much research interest, since their architecture changes the dominant paradigm in Computer Vision. An interesting and difficult task in this field is the classification of artwork styles, since the artistic style of a painting is a descriptor that captures rich information about the painting. In this paper, two different Deep Learning architectures—Vision Transformer and MLP Mixer (Multi-layer Perceptron Mixer)—are trained from scratch in the task of artwork style recognition, achieving over 39% prediction accuracy for 21 style classes on the WikiArt paintings dataset. In addition, a comparative study between the most common optimizers was conducted obtaining useful information for future studies.

Apical Amplification – A Cellular Mechanism of Conscious Perception?

We present a theoretical view of the cellular foundations fornetwork-level processes involved in producing our conscious experience.Inputs to apical synapses in layer 1 of a large subset of neocortical cellsare summed at an integration zone near the top of their apical trunk. Theseinputs come from diverse sources, and provide a context within which thetransmission of information abstracted from sensory input to their basal andperisomatic synapses can be amplified when relevant. We argue that apicalamplification (AA) enables conscious perceptual experience and makes it moreflexible, and thus more adaptive, by being sensitive to context. AA providesa possible mechanism for recurrent processing theory that avoids strongloops. It makes the broadcasting hypothesized by global neuronal workspacetheories feasible while preserving the distinct contributions of theindividual cells receiving the broadcast. It also provides mechanisms thatcontribute to the holistic aspects of integrated information theory. As AAis highly dependent on cholinergic, aminergic, and other neuromodulators, itrelates the specific contents of conscious experience to global mental statesand to fluctuations in arousal when awake. We conclude that apical dendritesprovide a cellular mechanism for the context-sensitive selectiveamplification that is a cardinal prerequisite of conscious perception.

V1 as an egocentric cognitive map

We typically distinguish between V1 as an egocentric perceptual map and the hippocampus as an allocentric cognitive map. In this article, we argue that V1 also functions as a post-perceptual egocentric cognitive map. We argue that three well-documented functions of V1, namely (i) the estimation of distance, (ii) the estimation of size, and (iii) multisensory integration, are better understood as post-perceptual cognitive inferences. This argument has two important implications. First, we argue that V1 must function as the neural correlates of the visual perception/cognition distinction and suggest how this can be accommodated by V1’s laminar structure. Second, we use this insight to propose a low-level account of visual consciousness in contrast to mid-level accounts (recurrent processing theory; integrated information theory) and higher-level accounts (higher-order thought; global workspace theory). Detection thresholds have been traditionally used to rule out such an approach, but we explain why it is a mistake to equate visibility (and therefore the presence/absence of visual experience) with detection thresholds.

Perception of invisible masked objects in early infancy

Recurrent loops in the visual cortex play a critical role in visual perception, which is likely not mediated by purely feed-forward pathways. However, the development of recurrent loops is poorly understood. The role of recurrent processing has been studied using visual backward masking, a perceptual phenomenon in which a visual stimulus is rendered invisible by a following mask, possibly because of the disruption of recurrent processing. Anatomical studies have reported that recurrent pathways are immature in early infancy. This raises the possibility that younger infants process visual information mainly in a feed-forward manner, and thus, they might be able to perceive visual stimuli that adults cannot see because of backward masking. Here, we show that infants under 7 mo of age are immune to visual backward masking and that masked stimuli remain visible to younger infants while older infants cannot perceive them. These results suggest that recurrent processing is immature in infants under 7 mo and that they are able to perceive objects even without recurrent processing. Our findings indicate that the algorithm for visual perception drastically changes in the second half of the first year of life.

The essential role of recurrent processing for figure-ground perception in mice

The segregation of figures from the background is an important step in visual perception. In primary visual cortex, figures evoke stronger activity than backgrounds during a delayed phase of the neuronal responses, but it is unknown how this figure-ground modulation (FGM) arises and whether it is necessary for perception. Here, we show, using optogenetic silencing in mice, that the delayed V1 response phase is necessary for figure-ground segregation. Neurons in higher visual areas also exhibit FGM and optogenetic silencing of higher areas reduced FGM in V1. In V1, figures elicited higher activity of vasoactive intestinal peptide–expressing (VIP) interneurons than the background, whereas figures suppressed somatostatin-positive interneurons, resulting in an increased activation of pyramidal cells. Optogenetic silencing of VIP neurons reduced FGM in V1, indicating that disinhibitory circuits contribute to FGM. Our results provide insight into how lower and higher areas of the visual cortex interact to shape visual perception.

The diachronic account of attentional selectivity

Many models of attention assume that attentional selection takes place at a specific moment in time which demarcates the critical transition from pre-attentive to attentive processing of sensory input. We argue that this intuitively appealing “synchronic” account is not only inaccurate, but has led to substantial conceptual confusion. As an alternative, we offer a “diachronic” framework that describes attentional selectivity as a process that unfolds over time. Key to this view is the concept of attentional episodes, brief periods of intense attentional amplification of sensory representations that regulate access to working memory and response-related processes. We describe how attentional episodes are linked to earlier attentional mechanisms and to recurrent processing at the neural level. We review studies that establish the existence of attentional episodes, delineate the factors that determine if and when they are triggered, and discuss the costs associated with processing multiple events within a single episode. Finally, we argue that this framework offers new solutions to old problems in attention research that have never been resolved. It can provide a unified and conceptually coherent account of the network of cognitive and neural processes that produce the goal-directed selectivity in perceptual processing that is commonly referred to as “attention”.

Perception of invisible masked objects in early infancy

Recurrent loops in the visual cortex play a critical role in visual perception, which is likely not mediated by purely feedforward pathways. However, the development of recurrent loops is poorly understood. The role of recurrent processing has been studied using visual backward masking, a perceptual phenomenon in which a visual stimulus is rendered invisible by a following mask, possibly because of the disruption of recurrent processing. Anatomical studies have reported that recurrent pathways are immature in early infancy. This raises the possibility that younger infants process visual information mainly in a feedforward manner, and thus, they might be able to perceive visual stimuli that adults cannot see because of backward masking. Here, we show that infants under 7 months of age are immune to visual backward masking and that masked stimuli remain visible to younger infants while older infants cannot perceive them. These results suggest that recurrent processing is immature in infants under 7 months and that they are able to perceive objects even without recurrent processing. Our findings indicate that the algorithm for visual perception drastically changes in the second half of the first year of life.

V1 as an Egocentric Cognitive Map

We typically distinguish between V1 as an egocentric perceptual map and the hippocampus as an allocentric cognitive map. In this article we explain why V1 also functions as an egocentric cognitive map. To the extent that cognitive processing has been discussed in V1, it has focused on (a) the allocation of attention, (b) top-down influences on perception, and (c) the transition from egocentric perception to allocentric navigation. By contrast, in this article we argue that three well-documented functions of V1, namely (a) the estimation of distance from eye position, (b) the estimation of size from eye position and/or pictorial cues, and (c) the multisensory integration of vision with proprioception and hearing, are potentially better understood as post-perceptual cognitive inferences. We use this insight to explore V1 as the neural correlates of the visual perception / cognition distinction, and propose a low-level account of visual consciousness in contrast to mid-level accounts (recurrent processing theory; integrated information theory), and higher-level accounts (higher-order thought; global workspace theory). We conclude by outlining the implications of our account for the perception of depth, motion, and colour / illumination.