翻译如图 5 所示尽管GIoU损失函数解决了真实边界框和预测边界框不相交的问题然而当预测框在真实框中时即 A∩B=B 的情况并不能很好地获取预测框的位置信息。

As shown in Figure 5, although the GIoU loss function solves the problem of non-intersecting between the true bounding box and the predicted bounding box, when the predicted box is within the true box, i.e., A∩B=B, it cannot obtain the position information of the predicted box well.