DensSiam Tracker
DensSiam: End-to-End Densely-Siamese Network with Self-Attention Model for Object Tracking
In this work, a novel convolutional Siamese architecture is proposed, which uses the concept of dense layers and connects each dense layer to all layers in a feed-forward fashion with a similarity-learning function. DensSiam also includes a Self-Attention mechanism to force the network to pay more attention to the non-local features during offline training. Extensive experiments are performed on four tracking benchmarks: OTB2013 and OTB2015 for vali- dation set; and VOT2015, VOT2016 and VOT2017 for testing set. The obtained results show that DensSiam achieves superior results on these benchmarks compared to other current state-of-the-art methods. DensSiam works beyond realtime at 60 FPS.
Experiments
We divided the benchmarks into two sets, the validation set which includes OTB2013, OTB2015 and the testing set which includes VOT2015 VOT2016 and VOT2017.
Visual results
In this section we show the results on VOT2016 and the overlapped sequences from VOT2015 and VOT2017 in three categories according to the official VOT website: easiest, intermediate and challenging.