Jiaming Han

@colorfully-ybf See mmdet/ops

https://github.com/facebookresearch/votenet/issues/2 Hope it works for you.

I guess you need to change the version of torch. We use torch==1.3 by default.

需要

Can you provide the log (console output)?

There should be `937`/`10833` testing images before/after spliting. Please check your data.

There are 2806 images. See our paper: DOTA: A Large-scale Dataset for Object Detection in Aerial Images

似乎cuda11是没法安装pytorch1.3的？

Assume we have feature maps F: [1,C,H,W], we just need a maxpooling on the channel dimension F_max = maxpool(F): [1,1,H,W]. Then we can use matplotlib to visualize F_max.

use `F.tensor` to get the pytorch tensor.