Busy life

The idea of SimpleShot is almost the same as that of ProtoNet

1

ProtoNet also uses the mean of each support class as its center and then implement Euclidean distance to train the network. The difference between SimpleShot and ProtoNet is that SimpleShot...

First of all, thanks for the interesting work. The experimental results are still competitive to the current SOTA works. However, the introduction makes me confused and doubt the results. For...

Questions about Resnet 12 backbone

4

Hello, I am reading your codes and there are a few questions i want to ask. 1. What is the purpose of the outputs of **second last layer**? I found...

About the performance of tieredImageNet

2

Dear author, I downloaded your 1-shot tieredImageNet pretrained model and implemented with the default settings (i.e., group=64). However, I can only get 66.80% by averaging 10,000 episodes. Any suggestions to...

Hello, thanks for sharing the codes. Can you explain how you got Equation 5? Shouldn't it be [cos(.,.)-log(exp(cos(.,.)))]?

Hi, in the paper, you mentioned that you used cosine and linear classifiers, but I cannot find any information in the testCos.py file. Could you tell me what kind of...

No train transformation in metatrain

2

Hi, did you forget to specify the train_augmentation in the train_metatrain_FewTRUE.py file? The tran_augmentation is None when the setname is 'train'.

As shown in the paper, we need to calculate the similarities of local descriptors from the support and query set. In this aspect, we definitely need the information of unlabeled...

Hi Could you upload the trained model for creating tieredImageNet? Any help is appreciated.