search

joe-siyuan-qiao / ViP-DeepLab

217 stars 26 forks source link

About SemanticKITTI-DVPS #4

Closed p-andra closed 3 years ago

p-andra commented 3 years ago

Thank you for your work!

Regarding the semantic images from SemanticKITTI-DVPS, they seem to be pretty sparse compared to the ones in the paper. Do you process the semantic image in order to obtain more dense segments, if so, how?

semantic

joe-siyuan-qiao commented 3 years ago

Thanks for the question. The paper shows densified annotations for visualization purpose only. For training and evaluation, we used sparse annotations as what you showed here.

p-andra commented 3 years ago

Thank you!