• Facets (new session)
• Description
• Metadata
• Settings
  • Rule: ActivityStreamsMapasEquivalentb3sb3sifpdbprdf-labelfacetshttp://dbpedia.org/resource/inference/rules/dbpedia#http://dbpedia.org/resource/inference/rules/opencyc#http://dbpedia.org/resource/inference/rules/umbel#http://dbpedia.org/resource/inference/rules/yago#http://dbpedia.org/schema/property_rules#http://www.ontologyportal.org/inference/rules/SUMO#http://www.ontologyportal.org/inference/rules/WordNet#http://www.w3.org/2002/07/owl#ldpoplwebskos-transurn:det:rdf:labelvirtrdf-labelvirtrdf-urlNone
    
  • Inverse Functional Properties: Disabled (fastest) Apply to subjects only Apply to objects only Apply to both subjects and objects
  • "Same As": Disabled (fastest) Apply to subjects only Apply to objects only Apply to both subjects and objects (recommended) Apply to subjects, objects and predicates (not recommended on big datasets) Apply to predicates only (special use cases only)

About: Video imprint (computer vision)     Goto   Sponge   NotDistinct   Permalink

An Entity of Type : owl:Thing, within Data Space : dbpedia.demo.openlinksw.com associated with source document(s)

Proposed as an extension of image epitomes in the field of video content analysis, video imprint is obtained by recasting video contents into a fixed-sized tensor representation regardless of video resolution or duration. Specifically, statistical characteristics are retained to some degrees so that common video recognition tasks can be carried out directly on such imprints, e.g., event retrieval, temporal action localization. It is claimed that both spatio-temporal interdependences are accounted for and redundancies are mitigated during the computation of video imprints.