A new hybrid ML method which combines ISA and SFA to gain both spatial and temporal features for action recognition. Reaches over 44% meanAP on Hollywood2 dataset.
jizhihang / isa_sfa Goto Github PK
View Code? Open in Web Editor NEWThis project forked from guangylegend/isa_sfa
A new hybrid ML method which combines ISA and SFA to gain both spatial and temporal features for action recognition. Reaches over 44% meanAP on Hollywood2 dataset.