search

vt-vl-lab / FGVC

[ECCV 2020] Flow-edge Guided Video Completion
Other
1.55k stars 263 forks source link

How can I train my own dataset #42

Closed a695649350 closed 3 years ago

a695649350 commented 3 years ago

I saw 3 pre-trained models. But I don't know how to train with my own dataset.

gaochen315 commented 3 years ago

Hi, you don't need to train the model for your dataset. Just set up the environment according to the instruction, and run this.

Let me know if you have any questions.

a695649350 commented 3 years ago

Hi, you don't need to train the model for your dataset. Just set up the environment according to the instruction, and run this.

Let me know if you have any questions.

Maybe I misunderstood the actual principle of this project, still thanks for the great job

a695649350 commented 3 years ago

Hi, you don't need to train the model for your dataset. Just set up the environment according to the instruction, and run this.

Let me know if you have any questions.

Sorry to interrupt again. I want to know how to train the deepfill weight? The image inpainting part use Deepfillv1, but it is hard for me to train a suitable weight file

gaochen315 commented 3 years ago

I didn't train Deepfillv1 myself. I use the pre-trained weight. Why do you need to train Deepfillv1 yourself? Do you have a specific dataset that requires fine-tuning?

a695649350 commented 3 years ago

I didn't train Deepfillv1 myself. I use the pre-trained weight. Why do you need to train Deepfillv1 yourself? Do you have a specific dataset that requires fine-tuning?

It seems like if the inpainting area which does not appear in the whole video will not be good as expected.

So I think this part may depend on the training dataset.

dragen1860 commented 1 year ago

@gaochen315 I can not find the training code? please help me, thank you.