np.pad用法

numpy.pad(array, pad_width, mode='constant', **kwargs)[source]

填充数组

x = np.random.randn(2,2)
np.pad(x,((1,0),(1,0)))
"""
array([[ 0.        ,  0.        ,  0.        ],
       [ 0.        ,  0.19829972,  0.11900865],
       [ 0.        , -0.67066229,  0.37756379]])
"""
x = np.random.randn(2,2)
np.pad(x,((0,0),(1,0)))
"""
array([[ 0.        ,  0.16003707,  0.87616892],
       [ 0.        ,  0.31563495, -2.02220122]])
"""

可以看到，在上面的例子中pad_width表示的是对应的axis轴在那个方向增加填充

pad_width[1]是在axis=1上增加填充

pad_width[1]是一个元组，表示在对应方向上应该填充多少，不填充时为（0，0）

更多的例子

x = np.random.randn(2,2)
np.pad(x,((1,1),(1,0)))
"""
array([[ 0.        ,  0.        ,  0.        ],
       [ 0.        , -0.30620401,  0.82797464],
       [ 0.        ,  0.23009474,  0.76201118],
       [ 0.        ,  0.        ,  0.        ]])
"""
x = np.random.randn(2,2)
np.pad(x,((1,1),(1,1)))
"""
array([[ 0.        ,  0.        ,  0.        ,  0.        ],
       [ 0.        , -0.22232814, -0.20075807,  0.        ],
       [ 0.        ,  0.18656139,  0.41005165,  0.        ],
       [ 0.        ,  0.        ,  0.        ,  0.        ]])
"""
x = np.random.randn(2,2,2)
np.pad(x,((0,0),(1,1),(1,1)))
"""
array([[[ 0.        ,  0.        ,  0.        ,  0.        ],
        [ 0.        , -0.44712856,  1.2245077 ,  0.        ],
        [ 0.        ,  0.40349164,  0.59357852,  0.        ],
        [ 0.        ,  0.        ,  0.        ,  0.        ]],

       [[ 0.        ,  0.        ,  0.        ,  0.        ],
        [ 0.        , -1.09491185,  0.16938243,  0.        ],
        [ 0.        ,  0.74055645, -0.9537006 ,  0.        ],
        [ 0.        ,  0.        ,  0.        ,  0.        ]]])
"""