Figure 1
From: Robust online learning based on siamese network for ship tracking
The architecture of SiamOL. The middle left is the Siamese sub-network for feature extraction. RPN sub-network lies in the middle right, which has two branches, one for classification and the other for regression. The top is UpdateNet for model update. The online classifier lies at the bottom, which consists of three parts: compression, dual attention mechanism, and filter. The classification fusion result is obtained by fusing the online classifier score map with the Siamese classification score map.
