How to modify the configuration file when using styleganv2 to generate a single-channel 800x800 SAR image? about mmagic HOT 3 CLOSED

@19990101lrk Sorry for late response. To support 800x800 image, you can modify config as follow:

# model
model = dict(
-  generator=dict(out_size=512),
-  discriminator=dict(in_size=512),
+  generator=dict(out_size=800),
+  discriminator=dict(in_size=800),
    ema_config=dict(
        type='ExponentialMovingAverage',
        interval=1,
        momentum=1. - (0.5**(32. / (ema_half_life * 1000.)))),
    loss_config=dict(
        r1_loss_weight=10. / 2. * d_reg_interval,
        r1_interval=d_reg_interval,
        norm_mode='HWC',
        g_reg_interval=g_reg_interval,
        g_reg_weight=2. * g_reg_interval,
        pl_batch_shrink=2))

To train with single-scale image, you can manually convert you image from single channel to three channel by repeat it at color dimension, e.g.

img = Image.open(img_path).convert('RGB')

from mmagic.

@19990101lrk Sorry for late response. To support 800x800 image, you can modify config as follow:

# model
model = dict(
-  generator=dict(out_size=512),
-  discriminator=dict(in_size=512),
+  generator=dict(out_size=800),
+  discriminator=dict(in_size=800),
    ema_config=dict(
        type='ExponentialMovingAverage',
        interval=1,
        momentum=1. - (0.5**(32. / (ema_half_life * 1000.)))),
    loss_config=dict(
        r1_loss_weight=10. / 2. * d_reg_interval,
        r1_interval=d_reg_interval,
        norm_mode='HWC',
        g_reg_interval=g_reg_interval,
        g_reg_weight=2. * g_reg_interval,
        pl_batch_shrink=2))

To train with single-scale image, you can manually convert you image from single channel to three channel by repeat it at color dimension, e.g.

img = Image.open(img_path).convert('RGB')

Thank you for your answer, I will try to modify it.

from mmagic.

Close due to no response. Please reopen if you need more help.

from mmagic.