Despite the tremendous success of deep learning in computer vision, models still fall behind humans in generalizing to new input distributions. Existing benchmarks do not investigate the specific failure points of models by analyzing performance under many controlled conditions. Our study systematically dissects where and why models struggle with contour integration - a hallmark of human vision – by designing an experiment that tests object recognition under various levels of object fragmentation. Humans (n=50) perform at high accuracy, even with few object contours present. This is in contrast to models which exhibit substantially lower sensitivity to increasing object contours, with most of the over 1,000 models we tested barely performing above chance. Only at very large scales ($\sim5B$ training dataset size) do models begin to approach human performance. Importantly, humans exhibit an integration bias - a preference towards recognizing objects made up of directional fragments over directionless fragments. We find that not only do models that share this property perform better at our task, but that this bias also increases with model training dataset size, and training models to exhibit contour integration leads to high shape bias. Taken together, our results suggest that contour integration is a hallmark of object vision that underlies object recognition performance, and may be a mechanism learned from data at scale.

Humans are able to segment images effortlessly without supervision using perceptual grouping. Here, we propose a counter-intuitive computational approach to solving unsupervised perceptual grouping and segmentation: that they arise *because* of neural noise, rather than in spite of it. We (1) mathematically demonstrate that under realistic assumptions, neural noise can be used to separate objects from each other; (2) that adding noise in a DNN enables the network to segment images even though it was never trained on any segmentation labels; and (3) that segmenting objects using noise results in segmentation performance that aligns with the perceptual grouping phenomena observed in humans, and is sample-efficient. We introduce the Good Gestalt (GG) datasets --- six datasets designed to specifically test perceptual grouping, and show that our DNN models reproduce many important phenomena in human perception, such as illusory contours, closure, continuity, proximity, and occlusion. Finally, we (4) show that our model improves performance on our GG datasets compared to other tested unsupervised models by $24.9$%. Together, our results suggest a novel unsupervised segmentation method requiring few assumptions, a new explanation for the formation of perceptual grouping, and a novel potential benefit of neural noise.

In crowding, perception of a target deteriorates in the presence of nearby flankers. Traditionally, it is thought that visual crowding obeys Bouma's law, i.e., all elements within a certain distance interfere with the target, and that adding more elements always leads to stronger crowding. Crowding is predominantly studied using sparse displays (a target surrounded by a few flankers). However, many studies have shown that this approach leads to wrong conclusions about human vision. Van der Burg and colleagues proposed a paradigm to measure crowding in dense displays using genetic algorithms. Displays were selected and combined over several generations to maximize human performance. In contrast to Bouma's law, only the target's nearest neighbours affected performance. Here, we tested various models to explain these results. We used the same genetic algorithm, but instead of selecting displays based on human performance we selected displays based on the model's outputs. We found that all models based on the traditional feedforward pooling framework of vision were unable to reproduce human behaviour. In contrast, all models involving a dedicated grouping stage explained the results successfully. We show how traditional models can be improved by adding a grouping stage.

Classic reinforcement learning (RL) theories cannot explain human behavior in the absence of external reward or when the environment changes. Here, we employ a deep sequential decision-making paradigm with sparse reward and abrupt environmental changes. To explain the behavior of human participants in these environments, we show that RL theories need to include surprise and novelty, each with a distinct role. While novelty drives exploration before the first encounter of a reward, surprise increases the rate of learning of a world-model as well as of model-free action-values. Even though the world-model is available for model-based RL, we find that human decisions are dominated by model-free action choices. The world-model is only marginally used for planning, but it is important to detect surprising events. Our theory predicts human action choices with high probability and allows us to dissociate surprise, novelty, and reward in EEG signals.

Classically, visual processing is described as a cascade of local feedforward computations. Feedforward Convolutional Neural Networks (ffCNNs) have shown how powerful such models can be. However, using visual crowding as a well-controlled challenge, we previously showed that no classic model of vision, including ffCNNs, can explain human global shape processing. Here, we show that Capsule Neural Networks (CapsNets), combining ffCNNs with recurrent grouping and segmentation, solve this challenge. We also show that ffCNNs and standard recurrent CNNs do not, suggesting that the grouping and segmentation capabilities of CapsNets are crucial. Furthermore, we provide psychophysical evidence that grouping and segmentation are implemented recurrently in humans, and show that CapsNets reproduce these results well. We discuss why recurrence seems needed to implement grouping and segmentation efficiently. Together, we provide mutually reinforcing psychophysical and computational evidence that a recurrent grouping and segmentation process is essential to understand the visual system and create better models that harness global shape computations.

Accumulating evidence indicates that the human brain copes with sensory uncertainty in accordance with Bayes' rule. However, it is unknown how humans make predictions when the generative model of the task at hand is described by uncertain parameters. Here, we tested whether and how humans take parameter uncertainty into account in a regression task. Participants extrapolated a parabola from a limited number of noisy points, shown on a computer screen. The quadratic parameter was drawn from a bimodal prior distribution. We tested whether human observers take full advantage of the given information, including the likelihood of the quadratic parameter value given the observed points and the quadratic parameter's prior distribution. We compared human performance with Bayesian regression, which is the (Bayes) optimal solution to this problem, and three sub-optimal models, which are simpler to compute. Our results show that, under our specific experimental conditions, humans behave in a way that is consistent with Bayesian regression. Moreover, our results support the hypothesis that humans generate responses in a manner consistent with probability matching rather than Bayesian decision theory.

In crowding, perception of an object deteriorates in the presence of nearby elements. Although crowding is a ubiquitous phenomenon, since elements are rarely seen in isolation, to date there exists no consensus on how to model it. Previous experiments showed that the global configuration of the entire stimulus must be taken into account. These findings rule out simple pooling or substitution models and favor models sensitive to global spatial aspects. In order to investigate how to incorporate global aspects into models, we tested a large number of models with a database of forty stimuli tailored for the global aspects of crowding. Our results show that incorporating grouping like components strongly improves model performance.

Decisions about noisy stimuli require evidence integration over time. Traditionally, evidence integration and decision making are described as a one-stage process: a decision is made when evidence for the presence of a stimulus crosses a threshold. Here, we show that one-stage models cannot explain psychophysical experiments on feature fusion, where two visual stimuli are presented in rapid succession. Paradoxically, the second stimulus biases decisions more strongly than the first one, contrary to predictions of one-stage models and intuition. We present a two-stage model where sensory information is integrated and buffered before it is fed into a drift diffusion process. The model is tested in a series of psychophysical experiments and explains both accuracy and reaction time distributions.

Geometry is closely linked to visual perception; yet, very little is known about the geometry of visual processing beyond early retinotopic organization. We present a variety of perceptual phenomena showing that a retinotopic representation is neither sufficient nor necessary to support form perception. We discuss the popular "object files" concept as a candidate for non-retinotopic representations and, based on its shortcomings, suggest future directions for research using local manifold representations. We suggest that these manifolds are created by the emergence of dynamic reference-frames that result from motion segmentation. We also suggest that the metric of these manifolds is based on relative motion vectors.

One of the fundamental and puzzling questions in vision research is how objects are segmented from their backgrounds and how object formation evolves in time. The recently discovered shine-through effect allows one to study object segmentation and object formation of a masked target depending on the spatiotemporal Gestalt of the masking stimulus (Herzog & Koch, 2001). In the shine-through effect, a vernier (two abutting lines) precedes a grating for a very short time. For small gratings, the vernier remains invisible while it regains visibility as a shine-through element for extended and homogeneous gratings. However, even subtle deviations from the homogeneity of the grating diminish or even abolish shine-through. At first glance, these results suggest that explanations of these effects have to rely on high-level Gestalt terminology such as homogeneity rather than on low-level properties such as luminance (Herzog, Fahle, & Koch, 2001). Here, we show that a simple neural network model of the Wilson-Cowan type qualitatively and quantitatively explains the basic effects in the shine-through paradigm, although the model does not contain any explicit, global Gestalt processing. Visibility of the target vernier corresponds to transient activation of neural populations resulting from the dynamics of local lateral interactions of excitatory and inhibitory layers of neural populations.