search

liux0614 / yolo_nano

Unofficial implementation of yolo nano
MIT License
254 stars 72 forks source link

预训练模型的问题 #4

Closed Zmjcc closed 5 years ago

Zmjcc commented 5 years ago

请问您在训练时如何解决没有预训练模型的问题？我用darknet框架训练时在没有预模型的情况下训练voc数据集loss并不会收敛。

tonyyang1995 commented 5 years ago

I think you can start with a small dataset first. It is a good way to debug for your code.