search

Yaoyi-Li / GCA-Matting

Official repository for Natural Image Matting via Guided Contextual Attention
MIT License
390 stars 62 forks source link

How to test high resolution imgs？ #16

Open Feywell opened 3 years ago

Feywell commented 3 years ago

I want to train a matting model by my own dataset. It will be used in high resolution imgs. Why is it not clear. How can I train my model to use in high resolution imgs. By the way, why the img size use 640?

Yaoyi-Li commented 3 years ago

The image size is not set to be 640 and I think there is no limitation on the training image size. You can upload the error information you encountered.

Feywell commented 3 years ago

Sorry. I mean，How to train a model for high resolution imgs. I want to use model for 2k resolution imgs

Yaoyi-Li commented 3 years ago

If you want to train on 2k images without any croping or resizing, it is a little difficult. In the training phase each 2k image may require more than 10G GPU memory. You can try to edit this line or codes in dataloader. https://github.com/Yaoyi-Li/GCA-Matting/blob/5a999be4b00bb482658fa7013a74a2b8acddc78e/utils/config.py#L45

javismiles commented 3 years ago

@Yaoyi-Li what is your recommended size for the images resolution as input to your network during training? please give me a range that you recommend, what is in your view the ideal size of images to input to your gca training, or the one that you used or that you recommend, thank you very much :)