search

Anymake / DRN_CVPR2020

Code and Dataset for CVPR2020 "Dynamic Refinement Network for Oriented and Densely Packed Object Detection"
GNU General Public License v2.0
328 stars 43 forks source link

the number of train epoch #14

Open sunpeng96 opened 3 years ago

sunpeng96 commented 3 years ago

thank you for your excellent work! I am using you FSM in my object. I have a question that whether your method need more epoch to fit data. As you mentioned in your article, in DOTA and HRSC2016, your need 140 epochs to train. thank for your reply

Panxjia commented 3 years ago

@sunpeng96 Thanks for your interest in our work. The training time(epoch) before and after adding the FSM is the same. The baseline method is also trained for 140 epochs. We did not conduct experiments to compare the performance when training different epochs.