patrickschlachter / osr-using-ics Goto Github PK

Hello, thanks your sharing.
Training a network for splitting the known classes by runing function " train_logits_cnn" in "main.py“. However, ”cnn_latent“ network model is saved for function " train_logits_cnnin" in "cnn_pool.py", which is not trained and " "cnn" network model is trained. But the model "cnn_latent" is used for splitting the known classes. So I dont know the meaning of function "train_logits_cnn".

hi, i am trying to reproduce the result of baccu that you put on the paper. unfortunately, i only get 0.8859 using model named 'modified_vgg' on mnist dataset,while it is 0.943 on the paper.i dont know if i should change model or something .the setting is same as the github project but only change the default dataset with mnist.