tf_unet icon indicating copy to clipboard operation
tf_unet copied to clipboard

Published 20 hours ago • verified publisher icon jakeret

test prediction

Open flighted opened this issue 6 years ago • 3 comments

Hello,I am confused with the test result,It all black images.And I make a normalization in data process.In the training process, the loss decreases to 16.It can't go down any more.Please give me some adivise.. Your help would be much appreciated. Thank you!

flighted avatar Apr 08 '18 11:04 flighted

Have you experimented with the network architecture (# of layers, # of filters)?

jakeret avatar Apr 08 '18 15:04 jakeret

Thanks for your rely.I use layer=4,feature_foot=16, epoch =20. The train dataset is 700. I have no idea of the set of layer, feature_root, training_iters and epoch. cr loss qq 20180409192946 Any idea of what's happening or any tips? Thanks in advance!

flighted avatar Apr 09 '18 11:04 flighted

The default value for batch size is 1 and for training_iters is 10. This means after only 20 epochs you have only used 200 of your input images. I would suggest you increase the training_iters

jakeret avatar Apr 17 '18 17:04 jakeret