Object Extraction of Tennis Video Based on Deep Learning

Table 4

Newly added layers and parameters to improve the HyperNet network.

	Input layer	Output layer
Category forecast	Conv4_3	Conv4_3_Conf
	FC7	FC7_Conf
	Conv8_2	Conv8_2_Conf
	Conv9_2	Conv9_2_Conf
Location prediction	Conv3_1	Conv3_1
	Conv4_3	Conv4_3_loc_pl
	FC7	FC_7_loc_pl
	Conv8_2	Conv8_2_loc_p
	Conv9_2	Conv9_2_loc_p
	FC7_loc	FC7_f