Activity Recognition based on a Magnitude-Orientation Stream Network

Caetano, C., de Melo, V. H. C., dos Santos, J. A., Schwartz, W. R.

URL

Proceedings: Conference on Graphics, Patterns and Images (SIBGRAPI)

Year: 2017

Peer Reviewed : Yes

Results: 
UCF101 /UCF101 Eval: 93.8% (According to the results, just using our Magnitude-Orientation Stream (MOS), we outperform many methods . In comparison with C3D, we outperform them by 5.3 p.p. using our temporal stream and 8.6 p.p. when combining it with Very Deep Two-Stream. This indicates that our magnitude orientation approach learns temporal information better than the approaches that perform 3D convolution operations directly. It is worth mentioning that we also improved the results achieved by the original two-stream by)
HMDB /HMDB Eval: 66.2% (When compared with neural network methods, we were able to outperform many methods using the proposed Magnitude Orientation Stream (MOS). Furthermore, we were able to outperform the original two-stream by 6:6 p.p. just using our temporal stream.)