lenalee103
commented
2 years ago Hello, have you check the loss, it can show some clues. My code is only used for the pavement aggregates, and it has a classification aside in the discriminator network. If you can Tests the discriminator and generator loss separately, it will tell where is the problem.---- Replied Message ***@***.***>Date03/07/2022 15:35 ***@***.***> ***@***.***>Subject[lenalee103/Aggregates-Database-based-on-ACGAN] About acwgan-gp (Issue #2)Thank you very much for open-sourcing your code！Have you tried your code on other data ? I have met some problems training my own data with it. In the first few hundred epochs, the true and false accuracy will fluctuate around 50%, but it will gradually decrease to 0.00%, and the network cannot converge to 50% accuracy again. Can you give me some advice?
—Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android.
You are receiving this because you are subscribed to this thread.Message ID: @.***>
[
{
@.***": "http://schema.org",
@.***": "EmailMessage",
"potentialAction": {
@.***": "ViewAction",
"target": "https://github.com/lenalee103/Aggregates-Database-based-on-ACGAN/issues/2",
"url": "https://github.com/lenalee103/Aggregates-Database-based-on-ACGAN/issues/2",
"name": "View Issue"
},
"description": "View this Issue on GitHub",
"publisher": {
@.***": "Organization",
"name": "GitHub",
"url": "https://github.com"
}
}
]
upgf
Thank you very much for open-sourcing your code！Have you tried your code on other data ? I have met some problems training my own data with it. In the first few hundred epochs, the true and false accuracy will fluctuate around 50%, but it will gradually decrease to 0.00%, and the network cannot converge to 50% accuracy again. Can you give me some advice?