Sports Video Classification Framework Using Enhanced Threshold Based Keyframe Selection Algorithm and Customized CNN on UCF101 and Sports1-M Dataset

Table 7

Comparison of various deep learning architectures with the proposed model for the Sports1-M dataset.

Reference	Architectures	Activation function	Optimizer	Batch size	Epochs	Learning rate	Training accuracy (%)	Testing accuracy (%)
[26]	AlexNet	Softmax	SGDM	10	15	0.0001	87.14	89.25
[26]	GoogleNet	Softmax	SGDM	15	15	0.0001	88.04	90.58
[27]	VGG16	ReLu	SGDM	128	60	0.001	96.01	90.91
[27]	VGG19	ReLu	SGDM	128	60	0.001	97.05	91.74
[27]	CNN	Softmax	Adam	128	50	0.001	89.11	90.73
Proposed model	Customized CNN	Softmax	SGDM	128	60	0.001	82.52	89.75