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

Predicting human scanpaths when exploring panoramic videos is a challengingtask due to the spherical geometry and the multimodality of the input, and theinherent uncertainty and diversity of the output. Most previous methods fail togive a complete treatment of these characteristics, and thus are prone toerrors. In this paper, we present a simple new criterion for scanpathprediction based on principles from lossy data compression. This criterionsuggests minimizing the expected code length of quantized scanpaths in atraining set, which corresponds to fitting a discrete conditional probabilitymodel via maximum likelihood. Specifically, the probability model isconditioned on two modalities: a viewport sequence as the deformation-reducedvisual input and a set of relative historical scanpaths projected ontorespective viewports as the aligned path input. The probability model isparameterized by a product of discretized Gaussian mixture models to capturethe uncertainty and the diversity of scanpaths from different users. Mostimportantly, the training of the probability model does not rely on thespecification of "ground-truth" scanpaths for imitation learning. We alsointroduce a proportional-integral-derivative (PID) controller-based sampler togenerate realistic human-like scanpaths from the learned probability model.Experimental results demonstrate that our method consistently produces betterquantitative scanpath results in terms of prediction accuracy (by comparing tothe assumed "ground-truths") and perceptual realism (through machinediscrimination) over a wide range of prediction horizons. We additionallyverify the perceptual realism improvement via a formal psychophysicalexperiment and the generalization improvement on several unseen panoramic videodatasets.

Scanpath Prediction in Panoramic Videos via Expected Code Length Minimization