Parnia Shokri

The problem is that I go from `[64, 128, 128, 256]` --> `[64, 128, 128]` and the network gets larger! This doesn't make sense since I reduce the number of...

Have you been able to reproduce this issue?

If you compile these two models (`filter_num` is the difference): ` model = models.transunet_2d((input_height, input_width, 3), filter_num=[32, 64, 128, 256], n_labels=n_classes, embed_dim=600, num_mlp=1000, num_heads=4, num_transformer=4, activation='ReLU', mlp_activation='GELU', output_activation='Sigmoid', batch_norm=True, pool=True,...

Hi I'm looking for the data as well. Did you receive it?

You can find a few examples of the testing set here: https://github.com/duggalrahul/MICCAI17_EndoVis_RoboSeg