• Corpus ID: 13339932

Human perception in computer vision

  title={Human perception in computer vision},
  author={Ron Dekel},
  • Ron Dekel
  • Published 17 January 2017
  • Computer Science, Biology
  • ArXiv
Computer vision has made remarkable progress in recent years. Deep neural network (DNN) models optimized to identify objects in images exhibit unprecedented task-trained accuracy and, remarkably, some generalization ability: new visual problems can now be solved more easily based on previous learning. Biological vision (learned in life and through evolution) is also accurate and general-purpose. Is it possible that these different learning regimes converge to similar problem-dependent optimal… 
Generalisation in humans and deep neural networks
The robustness of humans and current convolutional deep neural networks on object recognition under twelve different types of image degradations is compared and it is shown that DNNs trained directly on distorted images consistently surpass human performance on the exact distortion types they were trained on.
ImageNet-trained deep neural networks exhibit illusion-like response to the Scintillating grid
DNN representations of Scintillating grid visual illusion images in which white disks are perceived to be partially black are characterized, suggesting that nonmonotonicity in a deep computational representation is a potential test for illusion-like response geometry in DNN models.
Angular Visual Hardness
Recent convolutional neural networks (CNNs) have led to impressive performance but often suffer from poor calibration. They tend to be overconfident, with the model confidence not always reflecting
3D shape estimation in a constraint optimization neural network
Machine Foveation: An Application-Aware Compressive Sensing Framework
The design of Machine Foveation is presented–a general-purpose, application-aware compressive sensing related framework that uses a cascaded network architecture integrating an autoencoder and application network to determine the importance of each pixel.


Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition
These evaluations show that, unlike previous bio-inspired models, the latest DNNs rival the representational performance of IT cortex on this visual object recognition task and propose an extension of “kernel analysis” that measures the generalization accuracy as a function of representational complexity.
Do Computational Models Differ Systematically from Human Object Perception?
  • R. PramodS. Arun
  • Computer Science
    2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
  • 2016
Critical elements missing in computer vision algorithms are revealed and point to explicit encoding of these properties in higher visual areas in the brain.
Perceptual learning in Vision Research
  • D. Sagi
  • Psychology, Biology
    Vision Research
  • 2011
Deep Supervised, but Not Unsupervised, Models May Explain IT Cortical Representation
The results suggest that explaining IT requires computational features trained through supervised learning to emphasize the behaviorally important categorical divisions prominently reflected in IT.
A 3D shape inference model matches human visual object similarity judgments better than deep convolutional neural networks
Evidence is reported that CNNs are, in fact, not good models of human visual perception and it is shown that a 3D shape inference model explains human performance on an object shape similarity task better than CNNs.
Deep Neural Networks: A New Framework for Modeling Biological Vision and Brain Information Processing.
  • N. Kriegeskorte
  • Biology, Computer Science
    Annual review of vision science
  • 2015
This work states that biologically faithful feedforward and recurrent computational models of how biological brains perform high-level feats of intelligence, including vision, are entering an exciting new era.
Performance-optimized hierarchical models predict neural responses in higher visual cortex
This work uses computational techniques to identify a high-performing neural network model that matches human performance on challenging object categorization tasks and shows that performance optimization—applied in a biologically appropriate model class—can be used to build quantitative predictive models of neural processing.
A computational model for predicting local distortion visibility via convolutional neural network trainedon natural scenes
This paper presents a convolutional-neural-network-based (CNN-based) model to predict local distortion visibility in natural scenes and demonstrates that it can indeed succeed in this task and can more accurately predict thresholds than modern gain-control-based models.
Where practice makes perfect in texture discrimination: evidence for primary visual cortex plasticity.
  • A. KarniD. Sagi
  • Biology, Psychology
    Proceedings of the National Academy of Sciences of the United States of America
  • 1991
This work reports remarkable long-term learning in a simple texture discrimination task where learning is specific for retinal input and suggests that learning involves experience-dependent changes at a level of the visual system where monocularity and the retinotopic organization of thevisual input are still retained and where different orientations are processed separately.
How transferable are features in deep neural networks?
This paper quantifies the generality versus specificity of neurons in each layer of a deep convolutional neural network and reports a few surprising results, including that initializing a network with transferred features from almost any number of layers can produce a boost to generalization that lingers even after fine-tuning to the target dataset.