Problem with summay with GPU about enet

[cmdbug]

I'm very sorry. I only tested the pointrend function at that time, so I didn't use GPU.

[liyanping0317]

@cmdbug Thanks anyway. From my opinion, it seems that pointrend could be a commonly used method to improve the boundry segmentation result no matter which segmentation network is used. Do you this is possible?

[liyanping0317]

@cmdbug Sorry to bother again, i have some problem about channel dismatch, do you have any material or notes of how to use pointrend on other networks.

[cmdbug]

@cmdbug Thanks anyway. From my opinion, it seems that pointrend could be a commonly used method to improve the boundry segmentation result no matter which segmentation network is used. Do you this is possible?

It can improve the segmentation of detail parts. It can be applied to other segmentation models.

[cmdbug]

@cmdbug Sorry to bother again, i have some problem about channel dismatch, do you have any material or notes of how to use pointrend on other networks.

I tried it by reading the paper and some information on the Internet.