JXingZhao
JXingZhao
还有就是多少个epoch收敛呢,我看每3个epoch下降一次学习率,这样没必要迭代30个epoch吧
It is because that the image path is wrong. You can check the image path. 发件人:luoqiang10605 发送日期:2019-09-10 15:41:40 收件人:JXingZhao/EGNet 抄送人:Subscribed 主题:[JXingZhao/EGNet] run problem (#6) when I run the commend :python3...
yeah, you are right. 发件人:luoqiang10605 发送日期:2019-09-10 16:00:31 收件人:JXingZhao/EGNet 抄送人:JXingZhao ,Comment 主题:Re: [JXingZhao/EGNet] run problem (#6) I use the DUTS-TR-Image dataset.should I need use the DUTS-TE/DUTS-TE-Image/ dataset — You are receiving...
不是的,我们并没有固定预训练模型的参数。是所有的参数一起优化。 多任务的话,就是loss加起来,一起反向传播就可以了。 发件人:Belye 发送日期:2019-09-19 14:07:13 收件人:JXingZhao/EGNet 抄送人:Subscribed 主题:[JXingZhao/EGNet] 多个loss的优化问题 (#9) 您好! 深度学习初学者 请问下您在训练的时候是不是说固定了预训练模型的参数,然后对其他层进行优化,就是这句Adam(filter(lambda p: p.requires_grad, self.net_bone.parameters()), lr=self.lr_bone, weight_decay=p['wd'])代码,还有网络中存在多个loss进行优化的时候,是不是将所有loss加起来再反向传播? — You are receiving this because you are subscribed to this thread....
Hi, You need to make a empty dir first. From: GF Date: 2020-02-04 13:38:58 To: JXingZhao/EGNet Cc: Subscribed Subject: [JXingZhao/EGNet] Cannot save results on test images (#23) Hi, I used...
是这样的,因为在saliency 任务中,bn层不是特别关键,而且我们使用的batchsize也是1,所以就直接把bn层的参数在训练和测试的时候全部freeze了 发件人:daishengxuan 发送日期:2019-10-22 13:59:39 收件人:JXingZhao/EGNet 抄送人:Subscribed 主题:[JXingZhao/EGNet] 模型训练时,代码中为什么设置的self.net_bone.eval(),而不是self.net_bone.train() (#17) 作者您好,有一个问题我弄不明白。就是在solver.py文件中,在build_model()成员方法中为什么将模型的方式设置为self.net_bone.eval(),在其后训练时,好像没有让其变为train(),一般不应该是train()吗,希望抽一点时间回复,谢谢作者。 — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it...
This dataset is relatively early and not widely used. We ask the author of the dataset directly for the images, so it is correct. As for the 797 images used...
Dear sir/madam: 1. The test.lst only contains the left image, like this: ./RGB/RGBD_data_100.jpg. However, if u set the test.lst as ./nju2k/LR/000799_left.png ./depth/000799_left.jpg. It is also okay. 2. sal_lst means the...
I suggest that you can debug step by step. First, you can check whether the picture was successfully input into the network, and then check whether the input picture is...
Another possible reason is that the input results of the network are not scaled to 0-255. 发件人:zhangqiudan 发送日期:2019-08-08 17:16:59 收件人:JXingZhao/ContrastPrior 抄送人:JXingZhao ,Comment 主题:Re: [JXingZhao/ContrastPrior] About the test script (#2) Dear...