我自己写的类似tensorflow的conv2d函数对padding=SAME的场景无法通过测试

Question

#现象
自己实现的conv2d在padding=same的参数下无法通过与tensorflow内部conv2d函数的对比测试，但是对padding=valid的可以
#我的代码

def conv2d(input, filter, stride, padding):
    # batch x height x width x channels
    in_s = input.shape
    print("in_s:", in_s)
    # height x width x in_channels x out_channels
    f_s = filter.shape
    print("f_s:", f_s)

    assert len(in_s) == 4, 'input size rank 4 required!'
    assert len(f_s) == 4, 'filter size rank 4 required!'
    assert f_s[2] == in_s[3], 'intput channels not match filter channels.'
    assert f_s[0] >= stride and f_s[1] >= stride, 'filter should not be less than stride!'
    assert padding in [
        'SAME', 'VALID'], 'padding value[{0}] not allowded!!'.format(padding)

    if padding != 'VALID':
        # 提示： 关于SAME和VALID的区别，请参考：https://www.tensorflow.org/api_docs/python/tf/nn/convolution
        ##################
        # Your code here #
        # 计算前后补零的个数
        if f_s[0] % 2 == 0:  # 卷积核尺寸为偶数
            height_front_padding = (f_s[0] // 2) - 1
            height_end_padding = (f_s[0] // 2)
        else:  # 卷积核尺寸为奇数
            height_front_padding = height_end_padding = (f_s[0] - 1) // 2

        if f_s[1] % 2 == 0:  # 卷积核尺寸为偶数
            width_front_padding = (f_s[1] // 2) - 1
            width_end_padding = (f_s[1] // 2)
        else:  # 卷积核尺寸为奇数
            width_front_padding = width_end_padding = (f_s[1] - 1) // 2
        print("padding: ", height_front_padding, height_end_padding, width_front_padding, width_end_padding)
        # 开一个空间、容器
        new_height = in_s[1] + height_front_padding + height_end_padding
        new_width = in_s[2] + width_front_padding + width_end_padding
        temp = np.zeros((in_s[0], new_height, new_width, in_s[3]), dtype=np.float)
        # 开始补零
        for b in range(in_s[0]):
            for c in range(in_s[3]):
                for i in range(in_s[1]):
                    for j in range(in_s[2]):
                        temp[b][height_front_padding + i][width_front_padding + j][c] = input[b][i][j][c]
        ##################
        input = temp
        in_s = input.shape

    out_shape = (np.array(in_s[1: 3]) -
                 np.array(f_s[:2])) // stride + 1
    out_shape = np.concatenate([in_s[:1], out_shape, f_s[-1:]])
    output = np.zeros(out_shape)
    ##################
    # Your code here #
    # 开始做卷积
    for b in range(in_s[0]):
        for f in range(f_s[3]):
            for i in range(out_shape[1]):
                for j in range(out_shape[2]):
                    # 一次卷积乘的过程
                    output[b][i][j][f] = (input[b, i * stride:i * stride + f_s[0], j * stride:j * stride + f_s[1]] * filter[:, :, :, f]).sum()
    ##################
    return output

#测试代码

# 先定义个计算图用于运行tf
input_tensor = tf.placeholder(
    tf.float32, shape=[None, None, None, None], name='input')
filter_tensor = tf.placeholder(
    tf.float32, shape=[None, None, None, None], name='filter')

output_tensor1 = tf.nn.conv2d(
    input_tensor, filter_tensor, padding='SAME', strides=[1, 2, 2, 1])

output_tensor2 = tf.nn.conv2d(
    input_tensor, filter_tensor, padding='VALID', strides=[1, 3, 3, 1])

output_tensor3 = tf.nn.conv2d(
    input_tensor, filter_tensor, padding='SAME', strides=[1, 2, 2, 1])
try:
    final_score = 0  # 这个是最终得分

    filter = np.random.uniform(size=[5, 5, 3, 8])

    output = conv2d(img, filter, 2, 'SAME')


    with tf.Session() as sess:
        output_tf = sess.run(
            output_tensor3,
            feed_dict={
                input_tensor: img,
                filter_tensor: filter
            })
    print("mine: ", output[0, 1, :, 0])
    print("goal: ", output_tf[0, 1, :, 0])
    
    assert output.shape == output_tf.shape, 'shape mismatch [{}] vs [{}]'.format(
        output.shape, output_tf.shape)
    final_score += 20  # shape算对了得20分

    diff = np.mean(np.abs(output - output_tf))
    assert diff < 1e-5, 'value mismatch [{}]'.format(
        diff)  # 如果这一行没有报错的话，那么实现可以认为是正确的。
    final_score += 30  # 数值算对了得30分

    print('test 1 passed...')

    filter = np.random.uniform(size=[5, 5, 3, 8])

    output = conv2d(img, filter, 3, 'VALID')
    with tf.Session() as sess:
        output_tf = sess.run(
            output_tensor2,
            feed_dict={
                input_tensor: img,
                filter_tensor: filter
            })

    
    assert output.shape == output_tf.shape, 'shape mismatch [{}] vs [{}]'.format(
        output.shape, output_tf.shape)
    final_score += 20  # shape算对了得20分

    diff = np.mean(np.abs(output - output_tf))
    assert diff < 1e-5, 'value mismatch [{}]'.format(
        diff)  # 如果这一行没有报错的话，那么实现可以认为是正确的。
    final_score += 30  # 数值算对了得30分

    print('test 2 passed...')

except Exception as ex:
    print(ex)

print('Your final score:[{}]'.format(final_score))

#测试运行结果

Answer 1

你好，我是有问必答小助手，非常抱歉，本次您提出的有问必答问题，技术专家团超时未为您做出解答


本次提问扣除的有问必答次数，将会以问答VIP体验卡（1次有问必答机会、商城购买实体图书享受95折优惠）的形式为您补发到账户。


因为有问必答VIP体验卡有效期仅有1天，您在需要使用的时候【私信】联系我，我会为您补发。