Closed 972821054 closed 1 year ago
May I ask why you set the learning rate in the yaml in fulldata to 0.02, will there be any problems with this learning rate?
Because the average bounding boxes and image qualities of fulldata
is lower than GOT-10k
, only when the optimizer uses a larger iteration step size can the model converge quickly.
@ thanks!
May I ask why you set the learning rate in the yaml in fulldata to 0.02, will there be any problems with this learning rate?