Incremental Embedding Learning via Zero-Shot Translation,AAAI Conference on Artificial Intelligence (AAAI), 2021.
Requirments
All training and test are done in Pytorch framework.
Pytorch vesion: 0.4.1
Python version: 3.6
Datasets
We evaluate our methods in CUB-200-2011 and CIFAR100. (Note: CUB-200-2011 do not split the train set and test set in the original folder, the splited datasets can be download from this link according to the original provided train/test text file.)
Train and test
run CUB:
bash run_demo_cub.sh
run Cifar100:
bash run_demo_cifar.sh
Acknowledgements
Our code structure is inspired by SDC.