【AdaptiveAvgPool3d】pytorch教程

b站视频： https://www.bilibili.com/video/BV1va411D7Ua

1 torch.nn > AdaptiveAvgPool3d

1.1 相关资料

官方文档：https://pytorch.org/docs/stable/generated/torch.nn.AdaptiveAvgPool3d.html
参考资料：
https://runebook.dev/zh-CN/docs/pytorch/generated/torch.nn.adaptiveavgpool3d
Pytorch常用Layer深度理解：https://zhuanlan.zhihu.com/p/371167523

1.2 CLASS torch.nn.AdaptiveAvgPool3d(output_size)

Applies a 3D adaptive average pooling over an input signal composed of several input planes.
对由多个输入平面组成的输入信号进行三维自适应平均池化

The output is of size D x H x W, for any input size. The number of output features is equal to the number of input planes.
对于任何输入尺寸,输出的尺寸为D×H×W。输出特征的数量等于输入平面的数量。

1.3 Parameters 参数

output_size – the target output size of the form D x H x W. Can be a tuple (D, H, W) or a single number D for a cube D x D x D. D, H and W can be either a int, or None which means the size will be the same as that of the input.
output_size –目标输出大小，格式为D x H x W。可以是元组（D，H，W），也可以是多维数据集D xD x D的单个数字D。D，H和W可以是 int 或“ None ，这意味着大小将与输入的大小相同。

1.4 Shape

我从视频流的角度来对Shape进行解释
N表示batch_size、C代表channels、D是视频流的深度、H是每帧图像的高度，W是每帧图像的宽度

视频流，每帧都是一个或RGB或灰度图，且每帧的通道数都是一样的，假设为3，对于一 个固定的视频流，其应用Conv3d的输入大小应为(1,3, d, h, w), 其中1等于batch size, 3等于输入channels， d是视频流的深度，h是每帧图像的高度，w是每帧图像的宽度。

2 测试

2.1 可视化

【腾讯文档】AdaptiveAvgPool3d：https://docs.qq.com/slide/DWGlMU0ppa1RlcVJW

2.2 平台

这个例子我在极链AI平台测试：https://cloud.videojj.com/auth/register?inviter=18452&activityChannel=student_invite

2.3 代码

import torch
import torch.nn as nn

# target output size of 5x7x9
m = nn.AdaptiveAvgPool3d((5,7,9))
input = torch.randn(5, 64, 8, 9, 10)
output = m(input)

print("input.shape:",input.shape)
print("output.shape:",output.shape)

input.shape: torch.Size([5, 64, 8, 9, 10])
output.shape: torch.Size([5, 64, 5, 7, 9])

import torch
import torch.nn as nn

# target output size of 7x7x7 (cube)
m = nn.AdaptiveAvgPool3d(7)
input = torch.randn(5, 64, 10, 9, 8)
output = m(input)

print("input.shape:",input.shape)
print("output.shape:",output.shape)

input.shape: torch.Size([5, 64, 10, 9, 8])
output.shape: torch.Size([5, 64, 7, 7, 7])

import torch
import torch.nn as nn

# target output size of 7x9x8
m = nn.AdaptiveAvgPool3d((7, None, None))
input = torch.randn(1, 64, 10, 9, 8)
output = m(input)

print("input.shape:",input.shape)
print("output.shape:",output.shape)

input.shape: torch.Size([1, 64, 10, 9, 8])
output.shape: torch.Size([1, 64, 7, 9, 8])

3 paddle关于AdaptiveAvgPool3d

官网链接：https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/nn/AdaptiveAvgPool3D_cn.html