Conv2d layer and Convtranspose2d layer calculation

conv2d 와 convtranspose2d layer를 거치고 난 후, shape를 알 수 있는 공식을 적어본다. 

1. Conv2d:

output_size = (Input_size - kernel_size + 2padding_size) / stride + 1 

 

2. Max_pooling:

output_size = (Input_size - pooling_size) / stride + 1 

 

3. Convtranspose2 d:

output_size = stride_size(input_size - 1 ) + kernel_size - 2 padding_size