It would be super cool if you could make a step by step tutorial on how to train the model on new datasets.

[Christoph-Schuhmann]

It would be super cool if you could make a step by step tutorial on how to train the model on new datasets. :)

I would like to use your technique to sort unlabled images into different domains, which then could be used to train GANs to create images of these domains.

[wvangansbeke]

Hi @Christoph-Schuhmann,

Thank you for your interest. Sounds like an interesting idea!

Listing the most important steps off the top of my head here:

• To train on new datasets, you need to adapt your dataloader to the format we use in this repo. Have a look at the data folder and make sure your __getitem__ method is similar to ours.
• Don't forget to add the path to your dataset in utils/mypath.py
• Make config files (similar to the ones in the configs folder) for the pretext and clustering steps. I highly recommend using them.
• Follow the steps in the readme

The same question was asked in issue #8, but if I have more time later I can go into more detail.

Hope this helps for now.

[wvangansbeke]

If something is still unclear, feel free to ask. I will close this issue for now.