Isn't the learning rate of direct-pretraining for RetinaNet 0.08 ? Wrt the paper, 0.04 is the setting on the less data. If I'm mistaken, my apologies.
Isn't the learning rate of direct-pretraining for RetinaNet 0.08 ?
Wrt the paper, 0.04 is the setting on the less data.
If I'm mistaken, my apologies.