problem about odious.py

[WangZZJJ]

I want use odiou_3D loss to my own network, but why the loss computed by odious.py can be -inf? Besides, the prediction will be NaN after training very little time?

[Vegeta2020]

It may be caused the unexpected values produced by the model without pre-training, if you used the pre-trained model, it would be avoided.

[Eaphan]

It may be caused the unexpected values produced by the model without pre-training, if you used the pre-trained model, it would be avoided.

Have you thought about the reason why the loss can be -inf? @Vegeta2020 @WangZZJJ

[Vegeta2020]

As the non-pre-trained model may produce unexpected values like negative ones, they cannot generate normal boxs for IoU calculation.