English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In laboratory object recognition tasks based on undistorted photographs, bothadult humans and Deep Neural Networks (DNNs) perform close to ceiling. Unlikeadults', whose object recognition performance is robust against a wide range ofimage distortions, DNNs trained on standard ImageNet (1.3M images) performpoorly on distorted images. However, the last two years have seen impressivegains in DNN distortion robustness, predominantly achieved throughever-increasing large-scale datasets$\unicode{x2014}$orders of magnitude largerthan ImageNet. While this simple brute-force approach is very effective inachieving human-level robustness in DNNs, it raises the question of whetherhuman robustness, too, is simply due to extensive experience with (distorted)visual input during childhood and beyond. Here we investigate this question bycomparing the core object recognition performance of 146 children (aged4$\unicode{x2013}$15) against adults and against DNNs. We find, first, thatalready 4$\unicode{x2013}$6 year-olds showed remarkable robustness to imagedistortions and outperform DNNs trained on ImageNet. Second, we estimated thenumber of $\unicode{x201C}$images$\unicode{x201D}$ children have been exposedto during their lifetime. Compared to various DNNs, children's high robustnessrequires relatively little data. Third, when recognizing objectschildren$\unicode{x2014}$like adults but unlike DNNs$\unicode{x2014}$relyheavily on shape but not on texture cues. Together our results suggest that theremarkable robustness to distortions emerges early in the developmentaltrajectory of human object recognition and is unlikely the result of a mereaccumulation of experience with distorted visual input. Even though currentDNNs match human performance regarding robustness they seem to rely ondifferent and more data-hungry strategies to do so.

The developmental trajectory of object recognition robustness: children are like small adults but unlike big deep neural networks