Not able to reach more than 20% precision

fangpin / siamese-pytorch

Implementation of Siamese Networks for image one-shot learning by PyTorch, train and test model on dataset Omniglot

279 stars 59 forks source link

Not able to reach more than 20% precision #4

Closed Ameyapores closed 5 years ago

Ameyapores commented 5 years ago

Hey, Somehow, I am unable to achieve a precision of more than 20% with the same exact code. Could you have a look? https://github.com/Ameyapores/one_shot_learning/blob/master/notebook.ipynb

fangpin commented 5 years ago

There is a lack of reorganizing your training and testing folder in your code. We should keep that both of the two folders are consisting of training or testing classes (character) instead of alphabet. Refer make_dataset.py for help. @Ameyapores