Integrated Mediapipe with a CNN Model for Arabic Sign Language Recognition

Table 2

Summary of the proposed model.

Network layers	Output shape	Parameters
conv2d_1 (Conv2D)	(None, 240, 240, 16)	448
Max_pooling2d (max-pooling2)	(None, 80, 80, 16)	0
conv2d_2 (Conv2D)D)	(None, 80, 80, 32)	4,640
max_pooling2d_2 (max-pooling2D)	(None, 26, 26, 32)	0
conv2d_3 (Conv2D)	(None, 26, 26, 64)	18,496
max_pooling2d_3 (max-pooling2D)	(None, 8, 8, 64)	0
batch_normalization (BatchNormalization)	(None, 8, 8, 64)	256
flatten (Flatten)	(None, 4,096)	0
dropout (Dropout)	(None, 4,096)	0
dense (Dense)	(None, 256)	1,048,832
Dense_1 (Dense)	(None, 28)	7,196
Total parameters	1,079,868