TensorRT PoolingLayer

TensorRT PoolingLayer

IPoolingLayer在通道中实现池。支持的池类型有maximum、average和maximum average混合。             

Layer Description: 2D pooling层描述:二维池             

用2D滤波器计算a维张量a上的池,生成B维的张量B。B的维数取决于a的维数、窗口大小r、对称填充p和步长s,因此:

 

PoolingType::kMAX

Maximum over elements in window.

PoolingType::kAVERAGE

Average over elements in the window.

PoolingType::kMAX_AVERAGE_BLEND

Hybrid of maximum and average pooling. The results of the maximum pooling and the average pooling are combined with the blending factor as (1-blendFactor)*maximumPoolingResult + blendFactor*averagePoolingResult to yield the result. The blendFactor can be set to a value between 0 and 1.

默认情况下,平均池是在池窗口和填充的输入之间的重叠处执行的。如果exclusive参数设置为true,则在池窗口和未添加的输入之间的重叠区域执行平均池。             

Layer Description: 3D pooling图层说明:三维池             

使用3D过滤器计算a维度张量a上的池,以生成维度B的张量B。B的维度取决于a的维度、窗口大小r、对称填充p和跨距s,因此:

 

 Where func is defined by one of the pooling types t:

PoolingType::kMAX

Maximum over elements in window.

PoolingType::kAVERAGE

Average over elements in the window.

PoolingType::kMAX_AVERAGE_BLEND

Hybrid of maximum and average pooling. The results of the maximum pooling and the average pooling are combined with the blending factor as (1-blendFactor)*maximumPoolingResult + blendFactor*averagePoolingResult to yield the result. The blendFactor can be set to a value between 0 and 1.

默认情况下,平均池是在池窗口和填充的输入之间的重叠处执行的。如果exclusive参数设置为true,则在池窗口和未添加的输入之间的重叠区域执行平均池。             

条件和限制             

2D或3D由输入核维数的数量决定。对于2D池,输入和输出张量应该有3个或更多维。对于3D池,输入和输出张量应具有4个或更多维度。

人工智能芯片与自动驾驶
原文地址:https://www.cnblogs.com/wujianming-110117/p/13919877.html