pooling for blobs with num_axes()>4

[shaibagon]

Trying to work with Blobs with num_axes()>4, Pooling layer exits with the following error:

F0225 11:36:13.204582 23703 pooling_layer.cpp:81] Check failed: 4 == bottom[0]->num_axes() (4 vs. 5) Input must have 4 axes, corresponding to (num, channels, height, width)

This check: CHECK_EQ(4, bottom[0]->num_axes()) seems to safeguard pooling from working with "non-legacy" blobs.

Can this layer be extended (like conv layer) to handle N-D Blobs as well?

Thank you very much,
Shai

[shelhamer]

See #2442. It needs to be revived and rebased to have N-D pooling like our N-D convolution.

[shaibagon]

@shelhamer thank you for your reference. I'll look into it!