Open andupotorac opened 3 days ago
Hi, thank you for your question!
I'm not quite sure if I understand your question correctly, but our proposed model strongly depends on its training dataset. Specifically, we separately trained it on the two publicly available datasets RICO (based on UI layouts) and PubLayNet (based on the layout of academic papers). If you want to apply it to your own type of data, it would be necessary to re-train the model with your own data, so it can learn the corresponding layout patterns.
In terms of the components, there are in theory no restrictions, but you would likely need to adjust some hyperparameters, such as the number of categories.
From the demo it is not clear if this model could be used to generate smaller components also - think a quiz, a poll, a video player, etc - where it would be able to randomize multiple UIs if requested.