在《Fully Convolutional Networks for Semantic Segmentation》中提到的IU就是IoU。
mean IU:不同类别识别准确度的平均值,比如一张图片要识别2个物体,mean IU 就是这两个物体分别的IU的平均值
def label_accuracy_score(label_trues, label_preds, n_class):
"""Returns accuracy score evaluation result.
- overall accuracy
- mean accuracy
- mean IU
- fwavacc
"""
hist = np.zeros(<