Can I change the resize value in dataset transform function? about dexined

[xavysp]

Hi,
your value (image size) must be a multiple of 16. I have not implemented, yet, for an arbitrary size. For 1000, the input image should be 1008x1008, don´t worries, just change in:
'CLASSIC': {
'img_height':1008,
'img_width': 1008,
'test_list': None,
'train_list': None,
'data_dir': 'data', # mean_rgb
'yita': 0.5
},

and DexiNed will resize for you.

[xavysp]

Hi @xavysp,

Thank you for your work.

I have a doubt. In the 'def transform(self, img, gt)' inside datasets.py, the images are resized to 400 x 464. Is there any reason to choose this value? I have images of dimensions of 1000 x 1000. Can I replace this value with 1000 x 1000 for training and testing? Please advise.

By the way, I have commented this line 400 x 464. This was used just to evaluate the DexiNed speed.

[rgkannan676]

Thank you