Bug: item is boolean

[Jarrome]

in /workspace/sign/Box_Discretization_Network-master/maskrcnn_benchmark/structures/segmentation_mask.py:427

selected_polygons.append(self.polygons[i])

But i is always True

I guess this is a bug?

[Jarrome]

Now I find when I use torch1.4, the item is tensor of booleans.

But when turn to torch1.1, the item is tensor of uint8s.

Anyway, it seems the code function works fine.

conda install pytorch=1.0 torchvision=0/2 cudatoolkit=10.0 -c pytorch
cd Box_Discretization_Network
CUDA_HOME=/usr/local/cuda-10.0 python setup.py build develop