Qiaolezi
Results
2
issues of
Qiaolezi
I want to know how the reference box works,why they can locate the object to be detected precisely? That is important to me,thank you very much!
没有改动网络;二分类的任务训练;数据集包含2.4w个图像;batch_size=24,input_size=256×256; 问题:在第99个eopch时,total_loss突然为nan,在第101个epoch时,val_loss也变成了nan,如图所示   我查了下,具体有这几种情况:梯度爆炸、学习率过大、损失函数不合适等等,但是不知道是哪种原因。 请问老师您有何见解?