kwimage.algo._nms_backend.py_nms

Fast R-CNN Copyright (c) 2015 Microsoft Licensed under The MIT License [see LICENSE for details] Written by Ross Girshick

Module Contents

Functions

py_nms(np_ltrb, np_scores, thresh, bias=1)

Pure Python NMS baseline.

kwimage.algo._nms_backend.py_nms.py_nms(np_ltrb, np_scores, thresh, bias=1)

Pure Python NMS baseline.

References

https://github.com/rbgirshick/fast-rcnn/blob/master/lib/utils/nms.py

Example

>>> np_ltrb = np.array([
>>>     [0, 0, 100, 100],
>>>     [0, 0, 100, 100],
>>>     [100, 100, 10, 10],
>>>     [10, 10, 100, 100],
>>>     [50, 50, 100, 100],
>>>     [100, 100, 150, 101],
>>>     [120, 100, 180, 101],
>>>     [150, 100, 200, 101],
>>> ], dtype=np.float32)
>>> np_scores = np.linspace(0, 1, len(np_ltrb))
>>> thresh = 0.1
>>> bias = 0.0
>>> keep = sorted(py_nms(np_ltrb, np_scores, thresh, bias))
>>> print('keep = {!r}'.format(keep))
keep = [2, 4, 5, 7]

Example

>>> from kwimage.algo._nms_backend.py_nms import *  # NOQA
>>> np_ltrb = np.array([
>>>     [0, 0, 100, 100],
>>>     [100, 100, 10, 10],
>>>     [10, 10, 100, 100],
>>>     [50, 50, 100, 100],
>>> ], dtype=np.float32)
>>> np_scores = np.array([.1, .5, .9, .1])
>>> keep = list(py_nms(np_ltrb, np_scores, thresh=0.0, bias=1.0))
>>> print('keep@0.0 = {!r}'.format(keep))
>>> keep = list(py_nms(np_ltrb, np_scores, thresh=0.2, bias=1.0))
>>> print('keep@0.2 = {!r}'.format(keep))
>>> keep = list(py_nms(np_ltrb, np_scores, thresh=0.5, bias=1.0))
>>> print('keep@0.5 = {!r}'.format(keep))
>>> keep = list(py_nms(np_ltrb, np_scores, thresh=1.0, bias=1.0))
>>> print('keep@1.0 = {!r}'.format(keep))
keep@0.0 = [2, 1]
keep@0.2 = [2, 1]
keep@0.5 = [2, 1, 3]
keep@1.0 = [2, 1, 3, 0]