python中的⼀维卷积conv1d和⼆维卷积conv2d 先来看⼆维卷积conv2d
conv2d(input, filter, strides, padding, use_cudnn_on_gpu=True, data_format="NHWC", dilations=[1, 1, 1, 1], name=None)
"""Computes a 2-D convolution given 4-D `input` and `filter` tensors."""
给定4维的输⼊张量和滤波器张量来进⾏2维的卷积计算。
input：4维张量，形状：[batch, in_height, in_width, in_channels]
filter：滤波器（卷积核），4维张量，形状：[filter_height, filter_width, in_channels, out_channels]
strides：滤波器滑动窗⼝在input的每⼀维度上，每次要滑动的步长，是⼀个长度为4的⼀维张量。 return：该函数返回⼀个张量，其类型与input输⼊张量相同。
再看⼀维卷积conv1d，python中的⼀维卷积最终还是通过⼆维卷积实现的，先将输⼊张量和滤波器的维度扩展，再调⽤⼆维卷积conv2d 来实现。 def conv1d(value,filters, stride, padding, use_cudnn_on_gpu=None, data_format=None, name=None) """Computes a 1-D convolution given 3-D input and filter tensors."""
给定三维的输⼊张量和滤波器来进⾏1维卷积计算。
input：3维张量，形状shape和data_format有关：
（1）data_format = "NWC", shape = [batch, in_width, in_channels]
（2）data_format = "NCW", shape = [batch, in_channels, in_width]
filters：3维张量，shape = [filter_width, in_channels, out_channels],
stride：滤波器窗⼝移动的步长，为⼀个整数。
padding：与上⽂⼀致。
由conv1d源码可以看出，⼀维卷积的实现，是先对输⼊张量和filter扩展了⼀维，然后调⽤⼆维卷积进⾏运算的：
value = pand_dims(value, spatial_start_dim) # 输⼊张量高清中国viad>惜春小札
filters = pand_dims(filters, 0) # 滤波器
result = gen_v2d(
value,
filters,
strides,
padding,
use_cudnn_on_gpu=use_cudnn_on_gpu,
data_format=data_format)
return array_ops.squeeze(result, [spatial_start_dim])
下⾯为conv1d完整源码：
def conv1d(value,
rrkkk
filters,
stride,山西太原重型机器厂
padding,
use_cudnn_on_gpu=None,
data_format=None,
name=None):
with ops.name_scope(name, "conv1d", [value, filters]) as name:收容教养制度退出历史舞台
# Reshape the input tensor to [batch, 1, in_width, in_channels]
if data_format is None or data_format == "NHWC" or data_format == "NWC": data_format = "NHWC"
spatial_start_dim = 1
strides = [1, 1, stride, 1]
elif data_format == "NCHW" or data_format == "NCW":
data_format = "NCHW"
spatial_start_dim = 2
strides = [1, 1, 1, stride]
else:
raise ValueError("data_format must be \"NWC\" or \"NCW\".")
value = pand_dims(value, spatial_start_dim)
第三种温暖filters = pand_dims(filters, 0)
result = gen_v2d(
value,
filters,
strides,
padding,
use_cudnn_on_gpu=use_cudnn_on_gpu,
data_format=data_format)
return array_ops.squeeze(result, [spatial_start_dim])
豫公网安备41160202000603
站长QQ:729038198
关于我们
投诉建议