Polygonized Silhouettes and Polygon Coding Based Feature Representation for Human Action Recognition

The characteristics of human silhouette shape can be used for action recognition and classification. In this paper, a novel feature extraction method the silhouette-based classification actions in videos is proposed. proposed based on polygonization images coding. Since conventional generation methods do not satisfy integrity silhouettes, Yolact++ modified as generator. Our innovative approach masks are silhouettes to overcome problem. For purpose, new image form called Poly Silhouette (PoS), Polygonization (PoG) algorithm Polygon Coding (PoC) have been developed. step on, but similar curve polygonization. It fast, adaptable, accurate contour coordinates PoS images. PoCs were generated by projecting each edge vector from corner onto angular areas codes formed. These grouped into k-mers genetic algorithms features. guarantees that vectors equal length any video. Thus, no additional required dimensionality By using different k-mer lengths, accuracy versus computation time was analyzed depicted figures. developed tested HMDB51 & UCF101 datasets: SVM 20.98%, 1.63% k-NN 4.96%, 6.83%, respectively, significant improvements achieved.