search

TimoBolkart / TEMPEH

TEMPEH reconstructs 3D heads in dense semantic correspondence from calibrated multi-view images in about 0.3 seconds.
https://tempeh.is.tue.mpg.de
Other
116 stars 7 forks source link

Generalization ability #2

Closed HeadReaper-hc closed 1 year ago

HeadReaper-hc commented 1 year ago

Impressive work！Can I use your pre-train model to test my own data or I need to train it myself?

TimoBolkart commented 1 year ago

Thank you. You will need to train it on your own data.

Karthik-Ragunath commented 10 months ago

Sorry if the question is very basic, can you please give some reason why this is the case? Why can't we directly use the multi-view images we have from a particular subject to generate his 3D mesh with the pre-trained model?