目标检测之IoU(intersecton over union)标准

来源：互联网发布：淘宝电脑版首页登录编辑：程序博客网时间：2024/05/01 13:37

https://www.pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/

绿色代表ground truth box，红色代表predictive box.

ground truth box 需要手动标注，predictive box是你选用的模型计算出来的结果。

将两者放在一起用IOU来衡量我们的模型检测目标的有效性。

从下图你可以很直观的看到IoU的计算。

Figure 2: Computing the Intersection of Union is as simple as dividing the area of overlap between the bounding boxes by the area of union (thank you to the excellent Pittsburg HW4 assignment for the inspiration for this figure).

Figure 3: An example of computing Intersection over Unions for various bounding boxes.

从python编程的角度实现

defbb_intersection_over_union(boxA,boxB):

# determine the (x, y)-coordinates of the intersection rectangle

xA= max(boxA[0],boxB[0])

yA= max(boxA[1],boxB[1])

xB= min(boxA[2],boxB[2])

yB= min(boxA[3],boxB[3])

# compute the area of intersection rectangle

interArea= (xB- xA+ 1)* (yB- yA+ 1)

# compute the area of both the prediction and ground-truthrectangles

boxAArea= (boxA[2]- boxA[0]+ 1)* (boxA[3]- boxA[1]+ 1)

boxBArea= (boxB[2]- boxB[0]+ 1)* (boxB[3]- boxB[1]+ 1)

# compute the intersection over union by taking the intersection

# area and dividing it by the sum of prediction + ground-truth

# areas - the interesection area

iou= interArea/ float(boxAArea+ boxBArea- interArea)

# return the intersection over union value

returniou

Figure 5: Our goal is to evaluate the performs of our object detector by using Intersection of Union. Specifically, we want to measure the accuracy of the predicted bounding box (red) against the ground-truth (green).

Figure 6: Computing the Intersection of Union using Python.

Figure 7: A slightly better Intersection over Union score.

Figure 8: Deriving the Intersection of Union evaluation metric for object detection.

Figure 9: Measuring object detection performance using Intersection over Union.

以上这正图片来自IJCV论文“weakly supervised localization and learning with generic knowledge”

图中黄色代表groundtruthbox，红色代表false positive 示例，绿色框代表 true positive，最右端的是未检测到的object，是FN的示例。

图中的Corloc=TP/(TP+FP)=2/3=66%.它与IOU存在的差别在于一个是正确的示例个数之比以及一个是面积之比。

阅读全文

'); })();