What dataset did you use?

[dfrumkin]

Hello Dennis! What dataset did you use for training?

[dbpprt]

Mainly StyleGAN synthesized portraits, the supervisely person dataset and the aisegment dataset, however it is important to highlight that the model is not fully trained just for a couple of epochs to validate the approach. Synthesized data contains a lot of noisy labels e.g. mislabeled fine elements however overall it yielded interesting results.

[mln00b]

@dennisbappert how are you generating the high-quality masks for training for the StyleGAN portraits?

[hackkhai]

@dennisbappert If you can share the dataset with high quality masks that would really be helpfull. mail id: [email protected]