Patrick

Results 7 comments of Patrick

> > > > 我也有这个问题,作者的实现的差不多5个G,测试准确率有78.9, pytorch官方实现差不多1.4个G,测试集准确率也只有60%,,请问有看到问题在哪吗? > > > > 而且pytorch官方实现速度快很多 > > > > > > > > > 我自己用官方的ResNet-18(改了下第一层卷积核大小适应cifar数据集尺寸)在cifar100的准确率是65%~70%之间。 > > > 我也尝试完全照搬作者的实验和整个工程去在cifar100上训练resnet18,但准确率只有66%,完全达不到>70%的准确率 > > > 不知道有没有和我一样情况的?无法复现作者的准确率? >...

try change `correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)` to `correct_k = correct[:k].reshape(-1).float().sum(0, keepdim=True)` in utils.py line 48

> It should be work if you change the code in ./tf_extended/metrics.py Line 51. > "return variables.Variable" =>> "return variables.VariableV1" This works for me, too! Thanks a lot.

> 局代理,即类似终端命令之类的也会直接代 建议把这一段添加到 readme 里面, 只看名称真的不知道啥意思

Thanks @kaijianding Build from source works well

> Method for training deep unsupervised sailency detection model. Training module implements 4 training methodology 1. Real: The backbone network is trainined with Ground Truth object, loss = mean((pred_y_i, y_gt_i))...