TPA-GAN

TPA-GAN copied to clipboard

问题

2

我用自己的数据t1与t2模态大概500张，测试了您的模型，发现任务判别器，以及分类网络的训练时训练的准确度接近100，而训练时的测试准确率只有50左右，是什么原因呢？同时一般判别器需要保存模型吗？我看有文件夹

tantantan007

Hello, I ran the generator you provided, '190_GLoss7.2265_TrainACC0.9844_TestACC0.7933_TestSEN0.8611_TestSPE0.7405_TestAUC0.8852_G.pth', but it is not suitable for the code architecture of Generator_Pyconv357_Attention class, I see the model version in the 'test_TPA_GAN.py' is '297_GLoss94.6087_TrainACC1.0_TestACC1.0_TestSEN1.0_TestSPE1.0_TestAUC1.0_G.pth',...

QuJing785464

我不知道任务判别器什么时候放入，似乎我训练的[123/200] Generator_Loss: 69.2359 Train_ACC:1.0000 200/200 Test_ACC:0.7050 141/200 Test_SEN:0.6200 62/100 Test_SPE:0.7900 79/100 Test_AUC:0.7794 Time_Taken: 85.99107193946838 sec 124 [124/200] Generator_Loss: 65.9726 Train_ACC:1.0000 200/200 Test_ACC:0.6900 138/200 Test_SEN:0.6600 66/100 Test_SPE:0.7200 72/100 Test_AUC:0.7817 Time_Taken: 87.36247897148132...

tantantan007