search

lyndonzheng / TFill

[CVPR 2022]: Bridging Global Context Interactions for High-Fidelity Image Completion
MIT License
161 stars 13 forks source link

Training from Scatch #21

Open ArielleZhang opened 1 year ago

ArielleZhang commented 1 year ago

When you train the model from scratch, is it normal that the img_g is completely noise? When continue training on the pretrained model, the img_g is close to the org_input but when training from scratch (I am currently on iter 200,000) the img_g is just blue or yellow.

Hydrageno commented 1 year ago

When you train the model from scratch, is it normal that the img_g is completely noise? When continue training on the pretrained model, the img_g is close to the org_input but when training from scratch (I am currently on iter 200,000) the img_g is just blue or yellow.

Did you solve the problem? I've met the same issue before, after training the model from scratch, the output is completely noisy.

ArielleZhang commented 1 year ago

When you train the model from scratch, is it normal that the img_g is completely noise? When continue training on the pretrained model, the img_g is close to the org_input but when training from scratch (I am currently on iter 200,000) the img_g is just blue or yellow.

Did you solve the problem? I've met the same issue before, after training the model from scratch, the output is completely noisy.

I guess I will try to train for more iterations to see, but so far it is still noise.

JinxMan25 commented 1 year ago

@ArielleZhang were you able to figure out the issue?