Table 1.

Configuration of the network used in all our experiments.

	R(m)	G(m)S	G(m)C	T(m)	N(m)
m = 1	10 resolutions r ∈ {2i/4|i = 0, …, 9}	1 × 1 × 1	3 × 3 × 1 subsampling	Gabor filters of (20)	5 × 5 × 2 window
m = 2	same	1 × 1 × 1	Location and scale from which template is originally sampled	Randomly selected n × n × 4 templates, n ∈{4, 8, 12, 16}, with zero mean and unit norm	S% of image area and depth d in scale, where S ∈ {10, 20, 30}, d ∈ {3, 5, 7}

Unless otherwise noted, n × m × l means a spatial step of n × m and a step of l across scales.