Deep Learning-Based Football Player Detection in Videos

Table 1

Model details in terms of modules, layers and output dimensions.

Module	Layers	Output dimension
conv1	16 filters
	Max pool 2 d (2 × 2)	[w/2, h/2, 16]
	Residual connection
conv2	32 filters
	32 filters
	Max pool 2 d (2 × 2)	[w/4, h/4, 32]
	Residual connection
conv3	32 filters
	32 filters
	Max pool 2 d (2 × 2)	[w/8, h/8, 32]
conv4	64 filters
	64 filters
	Max pool 2 d (2 × 2)	[w/16, h/16, 64]
conv5	64 filters
	64 filters
	Max pool 2 d (2 × 2)	[w/32, h/32, 32]
1 × 1 conv1	32 filters	[w/16, h/16, 32]
Player classifier	32 filters
	2 filters
	Sigmoid	[w/16, h/16, 1]
Bounding box	32 filters
	4 filters	[w/16, h/16, 4]