Closed StarCharlie closed 1 year ago
The multi-label classification datasets can be accessed from their official websites VOC2007, MSCOCO2014 and NUS-WIDE. The raw images of NUS-WIDE can be accessed from https://pan.baidu.com/s/1Bj-7fdrZAvUJPqAKrUkbbQ (verification code: s6oj).
For the OpenImages dataset, we only use the localized narratives of its "V6" version. The ~130MB jsonl file can be downloaded from the official site.
Thank you a lot!!!!
Hello, I am also interested in reproducing your work, but I find it difficult to construct DATAROOT directory. Would you be kind enough to provide us DATAROOT directory directly or give us a guideline about where to download all your datasets? Thank you very much!