TestFilterDetections test_mini_batch fails sometimes on some TensorFlow versions

[willgraf]

On both TensorFlow version 1.11.0 and 1.12.0 the FilterDetections test case test_mini_batch will fail most of the time inside TravisCI. It will pass eventually if re-run enough.

On TensorFlow version 1.11.0, sometimes the test fails its self.assertAllEqual(actual_scores, expected_scores) test due to a correctness problem (the wrong entry is supprussed).

On TensorFlow version 1.12.0 sometimes the tests fails its self.assertAllEqual(actual_scores, expected_scores) test due to a NonMaxSuppression shape issue.

[willgraf]

I think this is caused by #218. However, when score_threshold is passed to the nms function, we see failures in tensorflow 1.8 only. This is likely duplicate, but will leave it for now.

[willgraf]

It looks like the successful tests were just lucky, the PR build had failures for 1.11 and 1.12 on both 2.7 and 3.5. All tests passed for python 3.7, but perhaps that is just lucky.

[willgraf]

This issue is NOT related to score_threshold.

---

TensorFlow 1.12 failure:

>       np.testing.assert_array_equal(a, b, err_msg="\n".join(msgs))
E       AssertionError: 
E       Arrays are not equal
E       
E       not equal where = (array([0]), array([0]))
E       not equal lhs = [0.9]
E       not equal rhs = [1.]
E       Mismatch: 0.167%
E       Max absolute difference: 0.10000002
E       Max relative difference: 0.10000002
E        x: array([[ 0.9, -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. ,
E               -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. ,
E               -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. ,...
E        y: array([[ 1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
E               -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
E               -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,...

TensorFlow 1.11 failure:

>       np.testing.assert_array_equal(a, b, err_msg=msg)
E       AssertionError: 
E       Arrays are not equal
E       
E       Mismatch: 0.167%
E       Max absolute difference: 0.10000002
E       Max relative difference: 0.10000002
E        x: array([[ 1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. ,
E               -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. ,
E               -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. , -1. ,...
E        y: array([[ 1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
E               -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,
E               -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1., -1.,...