Open Ingvar-Y opened 3 years ago
This information is indeed not available right away unfortunately. If you are using the datasets with torchmeta.datasets.helpers
, then this information can be found by looking at the Resize
transforms, although it is a bit hidden. For example, in Omniglot the images have size 28. Otherwise if you are not using torchmeta.datasets.helpers
, the image sizes are the default ones from the extracted dataset (e.g. 105 in Omniglot).
What are the image sizes in datasets? It would be nice to know since torchmeta tries to standardise few-shot learning datasets and this info isn't readily available from looking at sources.