Open ltfAutos opened 1 year ago
The usage of batch normalization and ReLU follows ResNet [10], with exception that PSMNet does not apply ReLU after summation.
Why don't we use the relu Activation function at last? Will the effect be better? Can you briefly explain the reason?
The usage of batch normalization and ReLU follows ResNet [10], with exception that PSMNet does not apply ReLU after summation.
Why don't we use the relu Activation function at last? Will the effect be better? Can you briefly explain the reason?