You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am confused about the generation of the following sections:
1、queries__queries_biost_few_x/ where x in [1, 5] (train queries from SVOX + pseudo-target domains)
2、queries_n5_dx/ where x in [1, 5] (just 5 images randomly sampled from queries of the target domain)
3、queries_biost_few_x/ where x in [1, 5] (val queries from pseudo-target domain)
Do you have detailed instructions for this part? How can this be achieved? How many images should there be in each part of the dataset? The more detailed, the better. I need your help^_^
二、
Could you please provide detailed instructions for training and testing, the current usage documentation does not seem very clear.
Looking forward to receiving your kind reply. Thank you very much :)
The text was updated successfully, but these errors were encountered:
Hello, @francescomontagna ,-I really appreciate for your work in AdAGeo ! Congratulations:)
I need help with two questions:
一、
Before training, I need to standardize the format of the dataset, as shown in the link below(https://github.com/valeriopaolicelli/adageo-pytorch/blob/fda/datasets_details.md).
I am confused about the generation of the following sections:
1、queries__queries_biost_few_x/ where x in [1, 5] (train queries from SVOX + pseudo-target domains)
2、queries_n5_dx/ where x in [1, 5] (just 5 images randomly sampled from queries of the target domain)
3、queries_biost_few_x/ where x in [1, 5] (val queries from pseudo-target domain)
Do you have detailed instructions for this part? How can this be achieved? How many images should there be in each part of the dataset? The more detailed, the better. I need your help^_^
二、
Could you please provide detailed instructions for training and testing, the current usage documentation does not seem very clear.
Looking forward to receiving your kind reply. Thank you very much :)
The text was updated successfully, but these errors were encountered: