about test about monoscene

[LittlePotatoChip]

and I find that the folder /trained_models is empty

[anhquancao]

Hello,
You need to download the pre-trained model here
https://github.com/cv-rits/MonoScene#pretrained-models

[LittlePotatoChip]

Hello, You need to download the pre-trained model here https://github.com/cv-rits/MonoScene#pretrained-models

So,should I retrain the model?

[anhquancao]

I don't understand. You only need to download the pre-trained checkpoints and put them in the folder /trained_models

[LittlePotatoChip]

I don't understand. You only need to download the pre-trained checkpoints and put them in the folder /trained_models

[anhquancao]

This is the evaluation results of the trained models

[anhquancao]

Ah, I understand what do you mean. This is the MonoScene trained model, you don't need to retrain it. This is different from the pre-trained EfficientNet which is the pretrained 2D backbone and is required to train MonoScene from scratch.

[LittlePotatoChip]

Ah, I understand what do you mean. This is the MonoScene trained model, you don't need to retrain it. This is different from the pre-trained EfficientNet which is the pretrained 2D backbone and is required to train MonoScene from scratch.

The whole pre-train model?

[anhquancao]

Yes

[LittlePotatoChip]

Yes

how to train a model by myself to Evaluate?Can the codes save the model files?

[anhquancao]

Yes, the checkpoints are saved in $KITTI_LOG/checkpoints

[LittlePotatoChip]

Yes, the checkpoints are saved in $KITTI_LOG/checkpoints

Can I copy the checkpoints file to the the folder /trained_models and rename it to evaluate directly?

[anhquancao]

Yes