Is it possible to retrain the model with my own images from the checkpoint weights you made available?
Closed this issue · 1 comments
vinicius-mariano commented
Hello!
I would like to train my own images, without need to train all the synth90k dataset again.
That is, training just the images I have, but not loosing the pre-training of synth90k dataset.
Is it possible? How?
MaybeShewill-CV commented
@vinicius-mariano Yep but you need to adjust the final fc layer which is related to the classification classes nums and reinitialize it yourself:)