Jan Zdenek
Jan Zdenek
I see. Thanks. I guess I'll drop the full validation then. I've tried playing a bit with the iterator but it always gets stuck somewhere. There seemed to be some...
I explained how to train with a dataset of a different number of classes in this issue. https://github.com/ajbrock/BigGAN-PyTorch/issues/43#issuecomment-559071781
@VincieD Do you get a mismatch error? It's caused by the embedding layer. The embedding layer in the pretrained model has a shape of (1000, 128), so you can't use...
Yes, it doesn't use moving average. It may slightly hurt the performance but the difference shouldn't be very big.
I have the same issue. I trained it for about 1000 epochs and I get about 3% precision and 25% recall... It detects some text boxes correctly, but there is...
I have reimplemented EAST in Keras using the original tensorflow implementation but I don't think I used L2 loss. Despite that, I was able to produce roughly the same performance...
I see. It's unfortunate that we can't reproduce the original results with this code. I wonder what is/are the causes(s). I guess I'll go back to my Keras code.
> if you know if i have to put something in my github for the license of the code please tell me. I put MIT license to this repo so...
It seems to be a compatibility issue between TF and Keras. Are you using TF v2? I only used this code in TF v1 so it doesn't include any features...
What are the specs of the machine you use? Particularly the GPU. And tell me your TF and Keras version too, please.