Residual Attentional Siamese Networkfor High Performance Online Visual Tracking

lastmod 2019-03-12created 2019-03-123 edits

Learning Attentions: Residual Attentional Siamese Networkfor High Performance Online Visual Tracking

Remarks

This is another variant of SiamFC. It mainly attempts to improve performance using Attention.

SiamFC does only offline training, not online training. As a result, it is relatively weak at adapting to changes when tracking a target in real time. This paper attempts to address this drawback with an Attention Mechanism.

Documentation

Learning Attentions - Residual Attentional Siamese Networkfor High Performance Online Visual Tracking

http://www.dcs.bbk.ac.uk/~sjmaybank/CVPR18RASTrackCameraV3.3.pdf

SiamFC series

See also

• Motion estimation

Favorite site

• OpenResearch.ai - (RASNet) Learning Attentions: Residual Attentional Siamese Network for High Performance Online Visual Tracking