deform_conv2d

torchvision.ops.deform_conv2d(input: Tensor, offset: Tensor, weight: Tensor, bias: Optional[Tensor] = None, stride: Tuple[int, int] = (1, 1), padding: Tuple[int, int] = (0, 0), dilation: Tuple[int, int] = (1, 1), mask: Optional[Tensor] = None) → Tensor

执行可变形卷积 v2，如 可变形卷积网络 v2：更可变形，更好的结果 中所述，如果 mask 不是 None，并且执行可变形卷积，如 可变形卷积网络 中所述，如果 mask 是 None。

参数：

• input (Tensor[batch_size, in_channels, in_height, in_width]) – 输入张量

• offset (Tensor[batch_size, 2 * offset_groups * kernel_height * kernel_width, out_height, out_width]) – 用于卷积核中每个位置的偏移量。

• weight (Tensor[out_channels, in_channels // groups, kernel_height, kernel_width]) – 卷积权重，分成大小为 (in_channels // groups) 的组

• bias (Tensor[out_channels]) – 形状为 (out_channels,) 的可选偏差。默认值：None

• stride (int 或 Tuple[int, int]) – 卷积中心之间的距离。默认值：1

• padding (int 或 Tuple[int, int]) – 每个图像周围填充的零的高度/宽度。默认值：0

• dilation (int 或 Tuple[int, int]) – 核元素之间的间距。默认值：1

• mask (Tensor[batch_size, offset_groups * kernel_height * kernel_width, out_height, out_width]) – 用于卷积核中每个位置的掩码。默认值：None

返回：

卷积的结果

返回类型：

Tensor[batch_sz, out_channels, out_h, out_w]

示例：

>>> input = torch.rand(4, 3, 10, 10)
>>> kh, kw = 3, 3
>>> weight = torch.rand(5, 3, kh, kw)
>>> # offset and mask should have the same spatial size as the output
>>> # of the convolution. In this case, for an input of 10, stride of 1
>>> # and kernel size of 3, without padding, the output size is 8
>>> offset = torch.rand(4, 2 * kh * kw, 8, 8)
>>> mask = torch.rand(4, kh * kw, 8, 8)
>>> out = deform_conv2d(input, offset, weight, mask=mask)
>>> print(out.shape)
>>> # returns
>>>  torch.Size([4, 5, 8, 8])