视频分割中的IOU计算问题(一维向量的IOU)

对视频的帧进行分割得到向量，举例说明

p = [1,1,0,0,1,1,0,0,0,1,2,2,2,3,3,3,4,4,5,5]
y = [1,1,1,1,0,0,0,2,2,2,2,2,3,3,3,4,4,5,5,5]

其中，p为预测帧的分类，y为真实数据帧的分类。

计算p和y的IOU，也就是一维向量的IOU。首先明白IOU的概念，IOU的计算公式如下。

看如下简图，

 很容易看出交集与并集的计算公式如下，

intersection = minimum(p_end, y_end) - maximum(p_start, y_start)  
union = maximum(p_end, y_end) - minimum(p_start[j], y_start)

 因此，可以求出IOU.

完整代码如下。

import numpy as np

p = [1,1,0,0,1,1,0,0,0,1,2,2,2,3,3,3,4,4,5,5]
y = [1,1,1,1,0,0,0,2,2,2,2,2,3,3,3,4,4,5,5,5]
def get_labels_start_end_time(frame_wise_labels, bg_class=["background"]):
    labels = []
    starts = []
    ends = []
    last_label = frame_wise_labels[0]
    if frame_wise_labels[0] not in bg_class:
        labels.append(frame_wise_labels[0])
        #如果标签第一帧不是"background"，labels[0]=第一帧的动作
        starts.append(0)
        #starts[0]=0
    for i in range(len(frame_wise_labels)):
        if frame_wise_labels[i] != last_label:
            if frame_wise_labels[i] not in bg_class:
                labels.append(frame_wise_labels[i])
                starts.append(i)
            if last_label not in bg_class:
                ends.append(i)
            last_label = frame_wise_labels[i]
            #labels是动作标签
            #starts是动作开始的位置
            #end是动作结束的位置
    if last_label not in bg_class:
        ends.append(i + 1)
    return labels, starts, ends
p_label, p_start,p_end = get_labels_start_end_time(p, bg_class=["background"])
y_label, y_start,y_end = get_labels_start_end_time(y, bg_class=["background"])
tp = 0
fp = 0
hits = np.zeros(len(y_label))
overlap = 0.1

for j in range(len(p_label)):
    min_end_py = np.minimum(p_end[j], y_end)
    max_start_py = np.maximum(p_start[j], y_start)
    max_end_py = np.maximum(p_end[j], y_end)
    min_start_py = np.minimum(p_start[j], y_start)
    intersection = np.minimum(p_end[j], y_end) - np.maximum(p_start[j], y_start)  # 交叉
    union = np.maximum(p_end[j], y_end) - np.minimum(p_start[j], y_start)
    IoU = (1.0 * intersection / union) * ([p_label[j] == y_label[x] for x in range(len(y_label))])