Giters

yaoyao-liu / meta-transfer-learning

TensorFlow and PyTorch implementation of "Meta-Transfer Learning for Few-Shot Learning" (CVPR2019)

Home Page:https://lyy.mpi-inf.mpg.de/mtl/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

some questions about performs

ljjcoder opened this issue · comments

commented

it is excellent work! I have some questions about your method. The performance produced by the code is 60.2 and 74.3. I notice that it produced by meta-batch. if i want to reproduce the result of using HT meta-batch, what should I do ?

commented

Hi @ljjcoder,

Thanks for your interest in our work.

The implementation of the HT meta-batch has not been included in the public repository. We plan to make it public later. If you hope to run experiments with HT meta-batch, you may implement it yourself according to our paper. We'll let you know if we add HT meta-batch into this repository.

If you hope for higher performance, you may try the PyTorch code with a deeper backbone (ResNet-25).

If you have further questions, feel free to add more comments or create a new issue.