用Python实现一个基本的NMS算法-马春杰杰

以下是一个简单的Python示例代码，展示了如何实现基本的NMS算法：

import numpy as np

def nms(bboxes, scores, threshold):
    """
    Non-maximum suppression algorithm.
    """
    x1 = bboxes[:, 0]
    y1 = bboxes[:, 1]
    x2 = bboxes[:, 2]
    y2 = bboxes[:, 3]

    areas = (x2 - x1 + 1) * (y2 - y1 + 1)
    order = scores.argsort()[::-1]

    keep = []
    while order.size > 0:
        idx = order[0]
        keep.append(idx)

        xx1 = np.maximum(x1[idx], x1[order[1:]])
        yy1 = np.maximum(y1[idx], y1[order[1:]])
        xx2 = np.minimum(x2[idx], x2[order[1:]])
        yy2 = np.minimum(y2[idx], y2[order[1:]])

        w = np.maximum(0.0, xx2 - xx1 + 1)
        h = np.maximum(0.0, yy2 - yy1 + 1)

        overlap = w * h / areas[order[1:]]
        inds = np.where(overlap <= threshold)[0]
        order = order[inds + 1]

    return keep

import numpy as np

def nms(bboxes, scores, threshold):

"""

Non-maximum suppression algorithm.

"""

x1 = bboxes[:, 0]

y1 = bboxes[:, 1]

x2 = bboxes[:, 2]

y2 = bboxes[:, 3]

areas = (x2 - x1 + 1) * (y2 - y1 + 1)

order = scores.argsort()[::-1]

keep = []

while order.size > 0:

idx = order[0]

keep.append(idx)

xx1 = np.maximum(x1[idx], x1[order[1:]])

yy1 = np.maximum(y1[idx], y1[order[1:]])

xx2 = np.minimum(x2[idx], x2[order[1:]])

yy2 = np.minimum(y2[idx], y2[order[1:]])

w = np.maximum(0.0, xx2 - xx1 + 1)

h = np.maximum(0.0, yy2 - yy1 + 1)

overlap = w * h / areas[order[1:]]

inds = np.where(overlap <= threshold)[0]

order = order[inds + 1]

return keep

其中，bboxes是一个n×4的数组，存储了n个候选框的坐标信息；scores是一个长度为n的数组，存储了每个候选框的置信度得分；threshold是重叠面积的阈值。

代码中的核心部分是while循环，它将每个候选框按照置信度从高到低排序，并遍历每个候选框，判断它是否需要被抑制。具体来说，它计算当前候选框与其余候选框的重叠面积，如果重叠面积大于阈值，则将该候选框抑制，否则保留该候选框。最终，函数返回保留下来的候选框的索引。这些候选框将被视为最终的检测结果，用于后续的处理或显示。