minuet.nn.functional.pooling

Functions

global_avg_pool(x)	Applies average pooling on the given SparseTensor.
global_max_pool(x)	Applies max pooling on the given SparseTensor.
global_sum_pool(x)	Applies sum pooling on the given SparseTensor.

global_avg_pool(x: SparseTensor)

Applies average pooling on the given SparseTensor. For each point cloud in the given SparseTensor, the feature tensor will be reduced with torch.mean() function. If there are multiple point clouds in the SparseTensor, the outputs will be stacked together with the batch order.

Parameters:

x – the given SparseTensor for average pooling

Returns:

the result tensor after average pooling

global_max_pool(x: SparseTensor)

Applies max pooling on the given SparseTensor. For each point cloud in the given SparseTensor, the feature tensor will be reduced with torch.max() function. If there are multiple point clouds in the SparseTensor, the outputs will be stacked together with the batch order.

Parameters:

x – the given SparseTensor for average pooling

Returns:

the result tensor after average pooling

global_sum_pool(x: SparseTensor)

Applies sum pooling on the given SparseTensor. For each point cloud in the given SparseTensor, the feature tensor will be reduced with torch.sum() function. If there are multiple point clouds in the SparseTensor, the outputs will be stacked together with the batch order.

Parameters:

x – the given SparseTensor for sum pooling

Returns:

the result tensor after sum pooling