from __future__ import division
from chainer.backends import cuda
[docs]def mask_iou(mask_a, mask_b):
"""Calculate the Intersection of Unions (IoUs) between masks.
IoU is calculated as a ratio of area of the intersection
and area of the union.
This function accepts both :obj:`numpy.ndarray` and :obj:`cupy.ndarray` as
inputs. Please note that both :obj:`mask_a` and :obj:`mask_b` need to be
same type.
The output is same type as the type of the inputs.
Args:
mask_a (array): An array whose shape is :math:`(N, H, W)`.
:math:`N` is the number of masks.
The dtype should be :obj:`numpy.bool`.
mask_b (array): An array similar to :obj:`mask_a`,
whose shape is :math:`(K, H, W)`.
The dtype should be :obj:`numpy.bool`.
Returns:
array:
An array whose shape is :math:`(N, K)`. \
An element at index :math:`(n, k)` contains IoUs between \
:math:`n` th mask in :obj:`mask_a` and :math:`k` th mask \
in :obj:`mask_b`.
"""
if mask_a.shape[1:] != mask_b.shape[1:]:
raise IndexError
xp = cuda.get_array_module(mask_a)
n_mask_a = len(mask_a)
n_mask_b = len(mask_b)
iou = xp.empty((n_mask_a, n_mask_b), dtype=xp.float32)
for n, m_a in enumerate(mask_a):
for k, m_b in enumerate(mask_b):
intersect = xp.bitwise_and(m_a, m_b).sum()
union = xp.bitwise_or(m_a, m_b).sum()
iou[n, k] = intersect / union
return iou