google-research / noisy-fewshot-learning Goto Github PK

Hi there,

Recently, I have a great interest in this work aiming at addressing such a few-shot-like task with additional noisy data.
Nevertheless, some problems about data pre-processing raised. As stated in this repository e.g. 13, 795 means that the 13th image of the YFCC100M dataset was used for the 795th class of Shot ImageNet Benchmark., I do not really understand what "13th" image exactly means. As I know, all images in YFCC100M are now stored in AWS s3 in a hierarchical manner, and there is no "image id" or something like that indicating "n-th" image.
May you have an instrument or clue to guide one to derive the "image id" for YFCC100M?

Hi,
thanks for sharing your implementation. I have two questions about it:

1. Does it also work on tabular data?
2. Is it possible to identify the noisy instances (return the noisy IDs or the clean set)?

Thanks!