search

facebookresearch / fewshotDatasetDesign

The paper studies the problem of learning to recognize a new class of objects from a very small number of labeled images. This is called few-shot learning. Previous work in the literature focused on designing new algorithms that allow to learn to generalize to new unseen classes.In this work, we consider the impact of the dataset that we train on, and experiment with some dataset manipulations to see which trade-offs are important in the design of a dataset aimed at few-shot learning.
Other
25 stars 6 forks source link

miniIN6K or miniIN1K #2

Open txw1997 opened 3 years ago

txw1997 commented 3 years ago

Hello, thanks for you excellent work! Where can I find the miniIN6K/miniIN1K cache file when compute the miniIN6K/miniIN1K? Can you provide the link to download?

indussky8 commented 3 years ago

Hello, thanks for you excellent work! Where can I find the miniIN6K/miniIN1K cache file when compute the miniIN6K/miniIN1K? Can you provide the link to download?

It is a great pity that all the authors have not replied our questions. BTW, have you reproduced the results or a part?

sbaio commented 3 years ago

Hello and thanks for your questions,

For the miniIN6k, it is resized down from IN6k. IN6k is constructed from IN22k. I provide the json file for all images kept in IN6k.

minin1k is exactly, the regular Imagenet with 1k classes but resized down to 84x84. Sorry I don't have the link to share with you.

You can find more details in the paper. Don't hesitate to ask if you still have questions about this.

Othman