Closed damonchang23 closed 8 months ago
Here are some related issues about training:
The Original ICON is trained on commercial datasets (e.g., RenderPeople), but the ICON in ECON's paper is re-trained on THuman2.0 to ease the comparison. You can refer to testing.md to see how to benchmark the methods.
First of all, thank you for sharing your great work.
I have a questions about training.
THuman2.0 has 525 human scans and as i know, in ICON, dataset categoried as 0-500(train), 501-505(test), 506-525(val). ECON also have a same partition to train the network?
Thank you again.