Temperature Net in PyTorch

We provide a PyTorch implementation of Temperature Net. The code contains some unnecessary test codes and will be re-organized soon.

Prerequisites

• Linux
• Python 3
• Pytorch 1.0
• GPU + CUDA CuDNN

Datasets

• miniImageNet.
• StanfordDog.
• StanfordCar.

Few-shot Classification

• For your convenience, we specify the parameters for every dataset, including the dataset path, # testing episodes, and # episodes to tune temperature (n*5/3). Please put the dataset in corresponding directory and directly run the corresponding file.
• Validation and test will be conducted while training.
• miniImageNet is splited as Ravi, Sachin, and Hugo Larochelle. "Optimization as a model for few-shot learning." (2016).
• Train a 5-way 1-shot model for miniImageNet:

python Train_5way1shot_miniImageNet.py 

• Train a 5-way 5-shot model for miniImageNet:

python Train_5way5shot_miniImageNet.py 

• Train a 5-way 1-shot model for StanfordDog:

python Train_5way1shot_StanfordDog.py 

• Train a 5-way 5-shot model for StanfordDog:

python Train_5way5shot_StanfordDog.py 

• Train a 5-way 1-shot model for StanfordCar:

python Train_5way1shot_StanfordCar.py 

• Train a 5-way 5-shot model for StanfordCar:

python Train_5way5shot_StanfordCar.py