Regarding the issue of datasets

[UnchangedOriInten]

What's the feature?

I would like to ask for the download address of the zero shot DOTA dataset, or which version of the DOTA dataset is used for partitioning, and what specific partitioning strategy is used, such as unlabeled data and labeled data in the training set? Also, I would like to ask how the DIOR dataset is used as an extension? Can you provide the code for the other three box selection strategies? Thank you.

Any other context?

No response

[lizzy8587]

Hi @UnchangedOriInten,

• we use DOTA-v1.0, the partition code cannot find now, if you need the final annotation files, pls email me.
• the code for other three box selection strategies have already included in the repo, just change some parameters in the semi_train_cfg: 1)semi_reg_uncs, semi_uncs_num; 2) semi_jitter_num, semi_select_jitter_num, and semie_reg_iter.

[UnchangedOriInten]

Thank you for your response.Yes,I need the final annotation files of DOTA-v1.0 and NWPU VHR-10.Which one is your Email?Thank you very much.