You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I find that there maybe a bug in the ROIAlign operator. For example,
x = mx.nd.array([[[[ 0., 1., 2., 3., 4., 5.],
[ 6., 7., 8., 9., 10., 11.],
[ 12., 13., 14., 15., 16., 17.],
[ 18., 19., 20., 21., 22., 23.],
[ 24., 25., 26., 27., 28., 29.],
[ 30., 31., 32., 33., 34., 35.],
[ 36., 37., 38., 39., 40., 41.],
[ 42., 43., 44., 45., 46., 47.]]]])
// region of interest i.e. bounding box coordinates.
y = mx.nd.array([[0,0,0,4,4]])
pooled_size = (2, 2)
spatial_scale = 1.0
The result of mx.nd.ROIAlign(x, y, pooled_size, spatial_scale) is
[[[[-3.4028235e+38 -3.4028235e+38]
[-3.4028235e+38 -3.4028235e+38]]]].
I find that there maybe a bug in the ROIAlign operator. For example,
x = mx.nd.array([[[[ 0., 1., 2., 3., 4., 5.],
[ 6., 7., 8., 9., 10., 11.],
[ 12., 13., 14., 15., 16., 17.],
[ 18., 19., 20., 21., 22., 23.],
[ 24., 25., 26., 27., 28., 29.],
[ 30., 31., 32., 33., 34., 35.],
[ 36., 37., 38., 39., 40., 41.],
[ 42., 43., 44., 45., 46., 47.]]]])
// region of interest i.e. bounding box coordinates.
y = mx.nd.array([[0,0,0,4,4]])
pooled_size = (2, 2)
spatial_scale = 1.0
The result of mx.nd.ROIAlign(x, y, pooled_size, spatial_scale) is
[[[[-3.4028235e+38 -3.4028235e+38]
[-3.4028235e+38 -3.4028235e+38]]]].
When I use ROIAlign from MxNet(https://github.com/apache/incubator-mxnet/blob/master/src/operator/contrib/roi_align.cc), I get the same result as caffe2's implementation of ROIAlign(https://github.com/pytorch/pytorch/blob/master/caffe2/operators/roi_align_op.cc).
The text was updated successfully, but these errors were encountered: