ethnhe / FFB6D

[CVPR2021 Oral] FFB6D: A Full Flow Bidirectional Fusion Network for 6D Pose Estimation.
MIT License
290 stars 72 forks source link

New model need to be trained with a new camera? #84

Open daihongxia opened 2 years ago

daihongxia commented 2 years ago

Hi @ethnhe, is that true that the model only works with the same camera? what if we have a new camera and we know its intrinsic matrix, do we still have to generate all the training data and retrain the model?