Annual Review of Vision Science最新文献

Since Lettvin and colleagues' seminal discovery of bug detector neurons in the frog retina, understanding how retinal circuits support behavioral demands has been a central goal of visual neuroscience. Recent advances in machine learning, genetic tools, and neural recording have transformed our understanding of these circuits, particularly in the mouse retina. With a focus on mice, we examine how species-specific visual sampling strategies determine the behavioral relevance of retinal computations and review recent insights into circuits underlying reflexive behaviors, threat detection, prey capture, color vision, and night vision. We also highlight how the behavioral state itself influences retinal processing through direct neuromodulation and pupillary changes, challenging the traditional view of purely feedforward retinal processing in mammals. These findings confirm the retina as a sophisticated computational engine whose circuits have evolved to meet species-specific behavioral demands. While Lettvin's discovery of dedicated retinal circuits for innate behaviors launched the field, new tools now promise to expand our understanding of retinal contributions to naturalistic and flexible behaviors across species.

Information visualization is central to how humans communicate. Designers produce visualizations to represent information about the world, and observers construct interpretations based on the visual input as well as their heuristics, biases, prior knowledge, and beliefs. Several layers of processing go into the design and interpretation of visualizations. This review focuses on processes that observers use for interpretation: perceiving visual features and their interrelations, mapping those visual features onto the concepts they represent, and comprehending information about the world based on observations from visualizations. Observers are more effective at interpreting visualizations when the design is well-aligned with the way their perceptual and cognitive systems naturally construct interpretations. By understanding how these systems work, it is possible to design visualizations that play to their strengths and thereby facilitate visual communication.

Over the past decade, artificial neural networks trained to classify images downloaded from the internet have achieved astounding, almost superhuman performance and have been suggested as possible models for human vision. In this article, we review experimental evidence from multiple studies elucidating the classification strategy learned by successful visual neural networks (VNNs) and how this strategy may be related to human vision as well as previous approaches to computer vision. The studies we review evaluate the performance of VNNs on carefully designed tasks that are meant to tease out the cues they use. The use of this method shows that VNNs are often fooled by image changes to which human object recognition is largely invariant (e.g., the change of a few pixels in the image or a change of the background or illumination), and, conversely, that the networks can be invariant to very large image manipulations that disrupt human performance (e.g., randomly permuting the patches of an image). Taken together, the evidence suggests that these networks have learned relatively low-level cues that are extremely effective at classifying internet images but are ineffective at classifying many other images that humans can classify effortlessly.

Visual scenes are often populated by densely layered and complex patterns of motion. The problem of motion parsing is to break down these patterns into simpler components that are meaningful for perception and action. Psychophysical evidence suggests that the brain decomposes motion patterns into a hierarchy of relative motion vectors. Recent computational models have shed light on the algorithmic and neural basis of this parsing strategy. We review these models and the experiments that were designed to test their predictions. Zooming out, we argue that hierarchical motion perception is a tractable model system for understanding how aspects of high-level cognition such as compositionality may be implemented in neural circuitry.

Layer 6 corticothalamic (L6 CT) pyramidal neurons send feedback projections from the primary visual cortex to both first- and higher-order visual thalamic nuclei. These projections provide direct excitation and indirect inhibition through thalamic interneurons and neurons in the thalamic reticular nucleus. Although the diversity of L6 CT pathways has long been recognized, emerging evidence suggests multiple subnetworks with distinct connectivity, inputs, gene expression gradients, and intrinsic properties. Here, we review the structure and function of L6 CT circuits in development, plasticity, visual processing, and behavior, considering computational perspectives on their functional roles. We focus on recent research in mice, where a rich arsenal of genetic and viral tools has advanced the circuit-level understanding of the multifaceted roles of L6 CT feedback in shaping visual thalamic activity.

Development of the vertebrate retina involves the interaction of multiple signaling pathways and cell types, and there is growing appreciation of the role of innate immune pathways in this process. Resident innate immune cells, particularly microglia, play myriad roles in retinal development, disease, and regeneration. Here we aim to highlight what is known about innate immune cell populations and pathways in retinal cell development and regeneration. Resident innate immune cells are present from the earliest stages of retinal development and regulate developmental cell elimination, synapse refinement, angiogenesis, and recovery from retinal damage. We discuss the signaling pathways mediating immune cell interactions with other cell populations in developing and regenerating retina and highlight species-specific differences in retinal innate immune cell function, which are particularly evident in retinal cell regeneration.

Over the past decade and a half, a new understanding has emerged of the role of vision during the critical period in the primary visual cortex. Rather than driving competition for cortical space, vision is now understood to inform the establishment of feature conjunctions that cannot be constructed intrinsically. Longitudinal imaging studies reveal that the establishment of these higher-order feature detectors is a remarkably dynamic process involving the gain and elimination of neurons from functional groups (e.g., binocular neurons with nonlinear response tuning). Experience exerts its influence selectively on this developing circuitry; some pathways require experience for normal development, while others appear to be intrinsically established. This difference drives the network dynamism that is exploited to construct novel cortical representations that best encode our local environment and inform our actions in it.

Artificial vision has advanced significantly on the basis of insights from human and animal vision. Still, biological vision retains advantages over mainstream computer vision, notably in terms of robustness, adaptability, power consumption, and compactness. Natural vision also demonstrates a great diversity of solutions to problems, adapted to specific tasks. Biological vision best corresponds to the subfield of computation imaging, in which optics and algorithms are codesigned to uncover scene information. We review current progress and opportunities in optics, sensors, algorithms, and joint designs that enable computational cameras to mimic the power of natural vision.

We review the current state of our knowledge of the neural control of vergence and ocular accommodation in primates including humans. We first describe the critical need for these behaviors for viewing in a three-dimensional world. We then consider the sensory stimuli that drive vergence eye movements and lens accommodation and describe models of the sensorimotor transformations required to drive these motor systems. We discuss the interaction of vergence with saccades to produce high-speed shifts in gaze between objects at different distances and eccentricities. We also cover the normal development of these eye movements as well as the sequelae associated with their maldevelopment. In particular, we examine the neural substrates that produce vergence and lens accommodation, including motoneurons, immediate premotor circuitry, cerebellar and precerebellar regions, and cerebral cortical areas.

Animal camouflage in the natural world has been studied for over a century, with early research often relying on descriptive accounts of patterning as perceived by human observers. Recent advances, however, have leveraged a deeper understanding of visual processing across a wide range of predators. This review examines literature illustrating how insights from vision science have enriched research on camouflage. We focus on three areas: color and texture, motion processing, and the perception of shape and depth. We discuss findings from vision research that show how animals seeking to remain undetected optimize their camouflage. We also explore how predator visual systems have evolved to break that camouflage. Last, we highlight gaps where vision science has yet to be applied to research on camouflage, with the hope of encouraging further interdisciplinary work.