search

czq142857 / BAE-NET

The code for paper "BAE-NET: Branched Autoencoder for Shape Co-Segmentation".
Other
68 stars 13 forks source link

Could the network be trained in class-agnostic manner? #5

Closed curiosity654 closed 3 years ago

curiosity654 commented 3 years ago

Hi Zhiqi, First, thank you for the amazing work. I find the models are trained with each class, unlike the full-supervised network that provides one network for all the classes. So I wonder could this network be trained in a class-agnostic manner?

czq142857 commented 3 years ago

Sadly, no. The network relies on the part-level correspondences between shapes in the training set. Therefore, it makes sense to train chairs and tables together (as shown in the paper), but chairs and airplanes do not make sense.