Pretrained weights about featdepth HOT 6 CLOSED

And what is the main difference between resume_from and finetune configs in your code?

from featdepth.

Hi! I'm curious about pretrained weights and configs: which config was used to receive this and which for this. I guess it cfg_kitti_fm.py and cfg_kitti_fm_refine.py respectively. Am I right?

Yes, you are right, the name of config corresponds with that of ckpt. Kitti refer to training data, fm refer to model, refine refer to online refinement.

from featdepth.

And what is the main difference between resume_from and finetune configs in your code?

The resume_from mode will reload pretrained weights and will continue the unfinished training.
If the training is finished, the resume_from mode will not continue training.
While finetune mode will training from the very beginning.

from featdepth.

And what is the main difference between resume_from and finetune configs in your code?

The resume_from mode will reload pretrained weights and will continue the unfinished training.
If the training is finished, the resume_from mode will not continue training.
While finetune mode will training from the very beginning.

So if I want to train pretrained on kitti weights with cityscapes dataset, I have to use finetune parameter? Or I can change total_epochs and data in config and set resume_from?

from featdepth.

The two ways will work the same, but I suggest you to use the finetune mode.

from featdepth.

Okay. Thank you a lot!

from featdepth.