Extracted data for FOL model training and testing

[dk-tran92]

Hello Mr. Yu Yao,

I have downloaded the features that you provided for TAD method (FOL model)
But I see that it don't include Ego-motion - the feature extracted from ORB-SLAM
I also found in your paper, table 3 shows Input of TAD is Box + Flow but not Egomotion; In the config Yaml, with_ego is set to False

Could you please add Ego-motion feature?
Or could you please confirm if you did not use it when you benchmarked TAD method on DoTA dataset?

[trThanhnguyen]

Hi, @dk-tran92, did you manage to train your model? Please share some of your experience. I'm wondering the same issue.