[英]Applying convolution operation to image - PyTorch
To render an image if shape 27x35 I use : 若要渲染图像(形状为27x35),请使用:
random_image = []
for x in range(1 , 946):
random_image.append(random.randint(0 , 255))
random_image_arr = np.array(random_image)
matplotlib.pyplot.imshow(random_image_arr.reshape(27 , 35))
This generates : 这会产生:
I then try to apply a convolution to the image using the torch.nn.Conv2d
: 然后,我尝试使用torch.nn.Conv2d
将卷积应用于图像:
conv2 = torch.nn.Conv2d(3, 18, kernel_size=3, stride=1, padding=1)
image_d = np.asarray(random_image_arr.reshape(27 , 35))
conv2(torch.from_numpy(image_d))
But this displays error : 但这显示错误:
~/.local/lib/python3.6/site-packages/torch/nn/modules/conv.py in forward(self, input)
299 def forward(self, input):
300 return F.conv2d(input, self.weight, self.bias, self.stride,
--> 301 self.padding, self.dilation, self.groups)
302
303
RuntimeError: input has less dimensions than expected
The shape of the input image_d
is (27, 35)
输入image_d
的形状为(27, 35)
image_d
(27, 35)
Should I change the parameters of Conv2d
in order to apply the convolution to the image ? 为了将卷积应用于图像,是否应该更改Conv2d
的参数?
Update. 更新。 From @McLawrence answer I have : 从@McLawrence答案我有:
random_image = []
for x in range(1 , 946):
random_image.append(random.randint(0 , 255))
random_image_arr = np.array(random_image)
matplotlib.pyplot.imshow(random_image_arr.reshape(27 , 35))
This renders image : 这将渲染图像:
Applying the convolution operation : 应用卷积运算:
conv2 = torch.nn.Conv2d(1, 18, kernel_size=3, stride=1, padding=1)
image_d = torch.FloatTensor(np.asarray(random_image_arr.reshape(1, 1, 27 , 35))).numpy()
fc = conv2(torch.from_numpy(image_d))
matplotlib.pyplot.imshow(fc[0][0].data.numpy()) matplotlib.pyplot.imshow(fc [0] [0] .data.numpy())
renders image : 渲染图像:
There are two problems with your code: 您的代码有两个问题:
First, 2d convolutions in pytorch
are defined only for 4d tensors. 首先,仅针对4d张量定义 pytorch
中的2d卷积。 This is convenient for use in neural networks. 这在神经网络中使用很方便。 The first dimension is the batch size while the second dimension are the channels (a RGB image for example has three channels). 第一维是批次大小,而第二维是通道(例如,RGB图像具有三个通道)。 So you have to reshape your tensor like 所以你必须像重塑你的张量
image_d = torch.FloatTensor(np.asarray(random_image_arr.reshape(1, 1, 27 , 35)))
The FloatTensor
is important here, since convolutions are not defined on the LongTensor
which will be created automatically if your numpy
array only includes int
s. FloatTensor
在这里很重要,因为在LongTensor
上未定义卷积,如果您的numpy
数组仅包含int
,则会自动创建卷积。
Secondly, You have created a convolution with three input channels, while your image has just one channel (it is greyscale). 其次,您创建了具有三个输入通道的卷积,而图像只有一个通道(灰度)。 So you have to adjust the convolution to: 因此,您必须将卷积调整为:
conv2 = torch.nn.Conv2d(1, 18, kernel_size=3, stride=1, padding=1)
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.