search

hustvl / ViTMatte

[Information Fusion (Vol.103, Mar. '24)] Boosting Image Matting with Pretrained Plain Vision Transformers
MIT License
371 stars 37 forks source link

training weight #8

Closed JWSunny closed 1 year ago

JWSunny commented 1 year ago

thanks for your sharing, the training model weight is 1.2G in my dataset，but provide the ViTMatte_xxx.pth is 337M, hwo to convert？

JWSunny commented 1 year ago

OK，i solved it~

JWSunny commented 1 year ago

the best model pth, how to confirm in training process?

JingfengYao commented 1 year ago

It is introduced in the train doc.

JWSunny commented 1 year ago

how to save the best model in test_data during training process? can i modify which file，not familiar with detectron2

JingfengYao commented 1 year ago

You could integrate evaluating during training by modifying https://github.com/hustvl/ViTMatte/blob/main/main.py#L41.

JWSunny commented 1 year ago

can you release the evaluating during training? How do you save the best model during training？

JingfengYao commented 1 year ago

We didn't do that. We just simply choose the weights from the last epoch of training as our final weights.

JWSunny commented 1 year ago

OK! i see, thanks~