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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Image-based sports analytics enable automatic retrieval of key events in agame to speed up the analytics process for human experts. However, mostexisting methods focus on structured television broadcast video datasets with astraight and fixed camera having minimum variability in the capturing pose. Inthis paper, we study the case of event detection in sports videos forunstructured environments with arbitrary camera angles. The transition fromstructured to unstructured video analysis produces multiple challenges that weaddress in our paper. Specifically, we identify and solve two major problems:unsupervised identification of players in an unstructured setting andgeneralization of the trained models to pose variations due to arbitraryshooting angles. For the first problem, we propose a temporal featureaggregation algorithm using person re-identification features to obtain highplayer retrieval precision by boosting a weak heuristic scoring method.Additionally, we propose a data augmentation technique, based on multi-modalimage translation model, to reduce bias in the appearance of training samples.Experimental evaluations show that our proposed method improves precision forplayer retrieval from 0.78 to 0.86 for obliquely angled videos. Additionally,we obtain an improvement in F1 score for rally detection in table tennis videosfrom 0.79 in case of global frame-level features to 0.89 using our proposedplayer-level features. Please see the supplementary video submission athttps://ibm.biz/BdzeZA.

Unsupervised Temporal Feature Aggregation for Event Detection in Unstructured Sports Videos