modified from https://github.com/open-mmlab/mmdetection Args: tcp_port: backend: Returns:

input: xyz: (m, 3), new_xyz: (n, 3), feat: (m, c), offset: (b), new_offset: (b) output: (n, c)

input: xyz: (m, 3), new_xyz: (n, 3), feat: (m, c), offset: (b), new_offset: (b) output: (n, c)

input: xyz: (m, 3), new_xyz: (n, 3), feat: (m, c), offset: (b of m), new_offset: (b of n) output: (n, c)

Array of IoU for each (non ignored) class

IoU for foreground class binary: 1 foreground, 0 background

input xyz: (n, 3), float input query_xyz: (m, 3), float input batch_idxs: (n), int input query_batch_offsets: (B + 1), int, offsets[-1] = m output idx

Binary Lovasz hinge loss logits: [B, H, W] Variable, logits at each pixel (between -\infty and +\infty) labels: [B, H, W] Tensor, bin

Monkey-patch an existing `DataParallel` object. Add the replication callback. Useful when you have customized `DataParallel` implementation.

input: xyz: (n, 3), new_xyz: (m, 3), feat: (n, c), idx: (m, nsample), offset: (b), new_offset: (b) output: new_feat: (m, c+3, nsample), group

input: xyz: (n, 3), new_xyz: (m, 3), feat: (n, c), idx: (m, nsample), offset: (b), new_offset: (b) output: new_feat: (m, nsample, c+3)

input: xyz: (n, 3), new_xyz: (m, 3), feat: (n, c), idx: (m, nsample), offset: (b), new_offset: (b) output: new_feat: (m, nsample, c+3), group

Helper function to synchronize (barrier) among all processes when using distributed training

Cross entropy loss