the mean dice is not good in acdc dataset
Wbbbjikj opened this issue · 7 comments
I trained the ACDC dataset using transunet, but the dice value is only 85. However, the original paper can reach 89. I don't know what the reason is.
I trained the ACDC dataset using transunet, but the dice value is only 85. However, the original paper can reach 89. I don't know what the reason is.
Did you changed any hyper parameter of TransUnet?
When I run TransUnet on ACDC dataset, loss don't decrease.
Yes, I tried many methods to adjust parameters, and finally found that only changing batchsize to 8 could improve the performance, but it could only stay at about 87 in the end. Did I adjust the parameters incomplete? Could you please provide your parameter Settings, thank you
I trained the ACDC dataset using transunet, but the dice value is only 85. However, the original paper can reach 89. I don't know what the reason is.
Did you changed any hyper parameter of TransUnet? When I run TransUnet on ACDC dataset, loss don't decrease.
Hello, can you provide me with the processed acdc data set? If you can provide me with a copy of the acdc data set, I will be very grateful. My email address is 1786225568@qq.com
Yes, I tried many methods to adjust parameters, and finally found that only changing batchsize to 8 could improve the performance, but it could only stay at about 87 in the end. Did I adjust the parameters incomplete? Could you please provide your parameter Settings, thank you
Hello, can you provide me with the processed acdc data set? If you can provide me with a copy of the acdc data set, I will be very grateful.My email address is 1786225568@qq.com
I trained the ACDC dataset using transunet, but the dice value is only 85. However, the original paper can reach 89. I don't know what the reason is.
Did you changed any hyper parameter of TransUnet? When I run TransUnet on ACDC dataset, loss don't decrease.
Hi ! it's so excited for me to see you have completed the experiment on ACDC wonderfully!!! and i also want to reproduce it,could you share your preprocessed ACDC dataset with me? 2863583613@qq.com thanks!
I trained the ACDC dataset using transunet, but the dice value is only 85. However, the original paper can reach 89. I don't know what the reason is.
Hello, can you provide me with the processed acdc data set? If you can provide me with a copy of the acdc data set, I will be very grateful. My email address is [1976939023@qq.com]
我使用transunet训练了ACDC数据集，但是dice值只有85，而原论文可以达到89，不知道是什么原因。
你好，我想请问一下你数据集是怎么划分的呀？70-10-20，一共有多少个切片呢