Dataset Information

The human visual system and CNNs can both support robust online translation tolerance following extreme displacements.

ABSTRACT: Visual translation tolerance refers to our capacity to recognize objects over a wide range of different retinal locations. Although translation is perhaps the simplest spatial transform that the visual system needs to cope with, the extent to which the human visual system can identify objects at previously unseen locations is unclear, with some studies reporting near complete invariance over 10 degrees and other reporting zero invariance at 4 degrees of visual angle. Similarly, there is confusion regarding the extent of translation tolerance in computational models of vision, as well as the degree of match between human and model performance. Here, we report a series of eye-tracking studies (total N = 70) demonstrating that novel objects trained at one retinal location can be recognized at high accuracy rates following translations up to 18 degrees. We also show that standard deep convolutional neural networks (DCNNs) support our findings when pretrained to classify another set of stimuli across a range of locations, or when a global average pooling (GAP) layer is added to produce larger receptive fields. Our findings provide a strong constraint for theories of human vision and help explain inconsistent findings previously reported with convolutional neural networks (CNNs).

SUBMITTER: Blything R

PROVIDER: S-EPMC7910631 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

The human visual system and CNNs can both support robust online translation tolerance following extreme displacements.

Blything Ryan R Biscione Valerio V Vankov Ivan I II Ludwig Casimir J H CJH Bowers Jeffrey S JS

Journal of vision 20210201 2

Visual translation tolerance refers to our capacity to recognize objects over a wide range of different retinal locations. Although translation is perhaps the simplest spatial transform that the visual system needs to cope with, the extent to which the human visual system can identify objects at previously unseen locations is unclear, with some studies reporting near complete invariance over 10 degrees and other reporting zero invariance at 4 degrees of visual angle. Similarly, there is confusio ...[more]

PMID: 33620380

Similar Datasets

Project description:BackgroundHealth research funding agencies are increasingly promoting evidence use in health practice and policy. Building on work suggesting how agencies can support such knowledge translation (KT), this paper discusses an online survey to assess KT training needs of researchers and research users as part of a Canadian provincial capacity-building effort.MethodsThe survey comprised 24 multiple choice and open-ended questions including demographics, interest in learning KT skills, likelihood of participating in training, and barriers and facilitators to doing KT at work. More than 1,200 people completed the survey. The high number of responses is attributed to an engagement strategy involving partner organizations (health authorities, research institutes, universities) in survey development and distribution. SPSS was used to analyze quantitative results according to respondents' primary role, geographic region, and work setting. Qualitative results were analyzed in NVivo.ResultsOver 85 percent of respondents are interested in learning more about the top KT skills identified. Research producers have higher interest in disseminating research results; research users are more interested in the application of research results. About one-half of respondents require beginner-level training in KT skills; one-quarter need advanced training. Time and cost constraints are the biggest barriers to participating in KT training. More than one-half of respondents have no financial support for travel and almost one-half lack support for registration fees. Time is the biggest challenge to integrating KT into work.ConclusionsOnline surveys are useful for determining knowledge translation training needs of researchers, research users and ultimately organizations. In this case, findings suggest the importance of considering all aspects of KT in training opportunities, while taking into account different stakeholder interests. Funders can play a role in developing new training opportunities as part of a broad effort, with partners, to build capacity for the use of health research evidence. Survey results would ideally be complemented with an objective needs assessment based on core competencies, and should be acted on in a way that acknowledges the complexity of knowledge translation in healthcare, existing training activities, and the expertise stakeholders already have but may not refer to as knowledge translation.

Dataset Information

The human visual system and CNNs can both support robust online translation tolerance following extreme displacements.

Publications

The human visual system and CNNs can both support robust online translation tolerance following extreme displacements.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure