About Train

[pokameng]

@chaofengc @guoshihuixmu
hello bros!!!
Thanks for great works!!!
I want to train Stage one and Stage two with both 512*512 images.
Can you tell me how to modify the configes [train_FeMaSR_HQ_pretrain_stage, train_FeMaSR_LQ_stage] ?

[chaofengc]

You might want to consider modifying the following codes:

FeMaSR/options/train_FeMaSR_HQ_pretrain_stage.yml

Line 18 in 1ca4295

gt_size: 256

FeMaSR/options/train_FeMaSR_HQ_pretrain_stage.yml

Line 43 in 1ca4295

gt_resolution: 256

Caution

The provided code is NOT optimized for $512\times512$ images. If you intend to work with this resolution, ensure you have a clear understanding of the required modifications. You will need to adjust other parts of the code accordingly to ensure proper functionality.

[pokameng]

You might want to consider modifying the following codes:

FeMaSR/options/train_FeMaSR_HQ_pretrain_stage.yml

Line 18 in 1ca4295

gt_size: 256

FeMaSR/options/train_FeMaSR_HQ_pretrain_stage.yml

Line 43 in 1ca4295

gt_resolution: 256

Caution

The provided code is NOT optimized for 512 × 512 images. If you intend to work with this resolution, ensure you have a clear understanding of the required modifications. You will need to adjust other parts of the code accordingly to ensure proper functionality.

okok！
So the pretrained models your hvae shared only support 256*256 images,right?
FeMaSR_HRP_model_g.pth
FeMaSR_HRP_model_d.pth

After I debug ,the vgg layer name should modify with "pool4" instead of "relu4_4"

I also email you by gmail!!!

[chaofengc]

Thank you for your email. I believe an open discussion here would be more beneficial, as it could also assist others who may have similar questions.

---

The pretrained model only supports $256 \times 256$ inputs, so you'll need to re-run both Stage 1 and Stage 2 when switching to $512 \times 512$. Be aware that additional issues may arise during this transition, so proceed with caution.

[pokameng]

Thank you for your email. I believe an open discussion here would be more beneficial, as it could also assist others who may have similar questions.

The pretrained model only supports 256 × 256 inputs, so you'll need to re-run both Stage 1 and Stage 2 when switching to 512 × 512 . Be aware that additional issues may arise during this transition, so proceed with caution.

Ok, thanks!!!