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

Knowledge Distillation (KD) is a strategy for the definition of a set oftransferability gangways to improve the efficiency of Convolutional NeuralNetworks. Feature-based Knowledge Distillation is a subfield of KD that relieson intermediate network representations, either unaltered or depth-reduced viamaximum activation maps, as the source knowledge. In this paper, we propose andanalyse the use of a 2D frequency transform of the activation maps beforetransferring them. We pose that\textemdash by using global image cues ratherthan pixel estimates, this strategy enhances knowledge transferability in taskssuch as scene recognition, defined by strong spatial and contextualrelationships between multiple and varied concepts. To validate the proposedmethod, an extensive evaluation of the state-of-the-art in scene recognition ispresented. Experimental results provide strong evidences that the proposedstrategy enables the student network to better focus on the relevant imageareas learnt by the teacher network, hence leading to better descriptivefeatures and higher transferred performance than every other state-of-the-artalternative. We publicly release the training and evaluation framework usedalong this paper athttp://www-vpu.eps.uam.es/publications/DCTBasedKDForSceneRecognition.

Attention-based Knowledge Distillation in Multi-attention Tasks: The Impact of a DCT-driven Loss