Could you provide the training file? about pns-net HOT 5 CLOSED

On Sep 8, 2021, at 10:29 AM, Daniel-Ji ***@***.***> wrote: Hi, Thanks for your attention to our work. We are planing to extend this work to journal version, and thus, the training code will be public in a near future. Besides, what's your problem encountered? Best regards, Ge-Peng. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#5 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AILOI52KL6P3NKB4NVIDMFTUA3C73ANCNFSM5DTT6XCQ>. Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

On Sep 8, 2021, at 10:55 AM, Daniel-Ji ***@***.***> wrote: Hi, We use this loss function in our implementation: class CrossEntropyLoss(nn.Module): def __init__(self): super(CrossEntropyLoss, self).__init__() def forward(self, *inputs): pred, target = tuple(inputs) total_loss = F.binary_cross_entropy(pred.squeeze(), target.squeeze().float()) return total_loss In my own experience, you can turn down the learning rate in the fine-tuning phase. The different tasks may incur different problems, but I think this core module could also work well in the video shadow segmentation task. Thus, you can make many attempts on tuning the hyper-parameters. If it still does not work, I recommend you can freeze or slow down the feature extraction layers. Best regards, Ge-Peng. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#5 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AILOI57RJW52AZL7HPW5TZTUA3GBRANCNFSM5DTT6XCQ>. Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.