search

dbpprt / u-2-net-portrait

The code for our newly accepted paper in Pattern Recognition 2020: "U^2-Net: Going Deeper with Nested U-Structure for Salient Object Detection."
Apache License 2.0
117 stars 14 forks source link

What dataset did you use? #6

Open dfrumkin opened 3 years ago

dfrumkin commented 3 years ago

Hello Dennis! What dataset did you use for training?

dbpprt commented 3 years ago

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 commented 3 years ago

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

hackkhai commented 3 years ago

@dennisbappert If you can share the dataset with high quality masks that would really be helpfull. mail id: khailashsanthakumar@gmail.com