Amharic Language Image Captions Generation Using Hybridized Attention-Based Deep Neural Networks

Table 1

Hyperparameters setup.

Hyperparameter	Values
Batch size	32
Embedding dimension	256
Dense layer	256
Hidden layer (unit)	512
Dropout for encoder and decoder	0.2 (for dense layer) and 0.2 (for GRU cell)
Recurrent dropout	0.2
Recurrent activation	ReLU
Kernel regularize and recurrent regularize	L2 (0.001) and L2 (0.1)
Optimizer	Adam
Epoch	35