Keras中的零填充是否需要存储空间?
model.add(ZeroPadding2D((1, 1)))
如果是的话,会是这样吗?
INPUT: [224x224x3] Memory: 224*224*3=150K Weight: 0
ZEROPAD_1: [226x226x3] Memory: 226*226*3=153K Weight: 0
CONV2D_1-64: [224x224x64] Memory: 224*224*64=3.2M Weight: ((3*3*3)*64)+64= 1,792