目标检测模型的性能评估

目标检测模型的性能评估

---

Reference：

1. https://blog.csdn.net/Katherine_hsr/article/details/79266880
2. https://towardsdatascience.com/what-is-map-understanding-the-statistic-of-choice-for-comparing-object-detection-models-1ea4f67a9dbd

---

1. 目标检测问题

Given an image, find the objects in it, locate their position and classify them.

2. Ground Truth

3. IOU

Intersection over Union is a ratio between the intersection and the union of the predicted boxes and the ground truth boxes.

3. MAP

MAP(mean average precision)is, literally, the average of all the average precisions(APs) of our classes in the dataset.

Precision = TP / (TP+FP)
举例：

1. IoU与阈值进行比较确认每个图像每个类的正确检测次数。使用IoU进行判断检测是否正确需要一个阈值（例如0.5），如果IoU>0.5, 则认为是真实的检测，反之则不是。
   即给定一张图像的类别C的Precision=图像正确预测(True Positives)的数量除以该图像中这一类的总的目标数量。
2. 一个C类的平均精度=在验证集上所有的图像对于类C的精度值的和/有类C这个目标的所有图像的数量。
3. MAP=所有类别的平均精度求和除以所有类别 。

转载请注明出处，谢谢！