python - 更改非最大抑制以选择Python中OpenCV的最小框

Question

一些快速谷歌搜索将发现您这种用于非最大抑制的代码。

import numpy as np

# Malisiewicz et al.
def non_max_suppression_fast(boxes, overlapThresh):
    if len(boxes) == 0:
        return []

    if boxes.dtype.kind == "i":
        boxes = boxes.astype("float")
    pick = []

    x1 = boxes[:,0]
    y1 = boxes[:,1]
    x2 = boxes[:,2]
    y2 = boxes[:,3]

    area = (x2 - x1 + 1) * (y2 - y1 + 1)
    idxs = np.argsort(y2)

    while len(idxs) > 0:
        last = len(idxs) - 1
        i = idxs[last]
        pick.append(i)

        xx1 = np.maximum(x1[i], x1[idxs[:last]])
        yy1 = np.maximum(y1[i], y1[idxs[:last]])
        xx2 = np.minimum(x2[i], x2[idxs[:last]])
        yy2 = np.minimum(y2[i], y2[idxs[:last]])

        w = np.maximum(0, xx2 - xx1 + 1)
        h = np.maximum(0, yy2 - yy1 + 1)

        overlap = (w * h) / area[idxs[:last]]

        idxs = np.delete(idxs, np.concatenate(([last],
            np.where(overlap > overlapThresh)[0])))

    return boxes[pick].astype("int")

这将为一组重叠的框返回最大的框，这些框的面积按功能中可以指示的百分比重叠。

您如何更改此设置以返回一组重叠的框中最小的框？

Answer 1

我认为以上代码是错误的。可能是您在复制代码时犯了一个小错误。但是，通过这个小小的改变就可以解决问题。
idxs = np.argsort(area)
要获得最小值，您只需要在上面的代码中进行以下更改:

# Malisiewicz et al.
def non_max_suppression_fast(boxes, overlapThresh):
  if len(boxes) == 0:
    return []

  if boxes.dtype.kind == "i":
    boxes = boxes.astype("float")
  pick = []

  x1 = boxes[:,0]
  y1 = boxes[:,1]
  x2 = boxes[:,2]
  y2 = boxes[:,3]

  area = (x2 - x1 + 1) * (y2 - y1 + 1)
  idxs = np.argsort(area)

  while len(idxs) > 0:
    last = len(idxs) - 1
    i = idxs[last]

    xx1 = np.maximum(x1[i], x1[idxs[:last]])
    yy1 = np.maximum(y1[i], y1[idxs[:last]])
    xx2 = np.minimum(x2[i], x2[idxs[:last]])
    yy2 = np.minimum(y2[i], y2[idxs[:last]])

    w = np.maximum(0, xx2 - xx1 + 1)
    h = np.maximum(0, yy2 - yy1 + 1)

    overlap = (w * h) / area[idxs[:last]]
    ### Modification
    selected_idx = np.where(overlap > overlapThresh)[0]
    selected_idx = np.concatenate(([last], selected_idx))
    min_area_idx = min(selected_idx, key=lambda i: area[i])
    pick.append(min_area_idx)
    ### Modification end
    idxs = np.delete(idxs, selected_idx)

 return boxes[pick].astype("int")