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

The core of everyday tasks like reading and driving is active objectrecognition. Attempts to model such tasks are currently stymied by theinability to incorporate time. People show a flexible tradeoff between speedand accuracy and this tradeoff is a crucial human skill. Deep neural networkshave emerged as promising candidates for predicting peak human objectrecognition performance and neural activity. However, modeling the temporaldimension i.e., the speed-accuracy tradeoff (SAT), is essential for them toserve as useful computational models for how humans recognize objects. To thisend, we here present the first large-scale (148 observers, 4 neural networks, 8tasks) dataset of the speed-accuracy tradeoff (SAT) in recognizing ImageNetimages. In each human trial, a beep, indicating the desired reaction time,sounds at a fixed delay after the image is presented, and observer's responsecounts only if it occurs near the time of the beep. In a series of blocks, wetest many beep latencies, i.e., reaction times. We observe that human accuracyincreases with reaction time and proceed to compare its characteristics withthe behavior of several dynamic neural networks that are capable ofinference-time adaptive computation. Using FLOPs as an analog for reactiontime, we compare networks with humans on curve-fit error, category-wisecorrelation, and curve steepness, and conclude that cascaded dynamic neuralnetworks are a promising model of human reaction time in object recognitiontasks.

SATBench: Benchmarking the speed-accuracy tradeoff in object recognition by humans and dynamic neural networks