Fig. 3From: Spatio-temporal information for human action recognitionScreenshots of KTH datasetBack to article page