search

YoungXIAO13 / PoseFromShape

(BMVC 2019) PyTorch implementation of Paper "Pose from Shape: Deep Pose Estimation for Arbitrary 3D Objects"
http://imagine.enpc.fr/~xiaoy/PoseFromShape/
MIT License
170 stars 34 forks source link

using GT boxes to crop images #15

Closed kartikgupta-at-anu closed 4 years ago

kartikgupta-at-anu commented 4 years ago

Are you using GT boxes to crop the images in all the experiments in this paper? I am assuming StarMap, as well as all other baselines, do not use the cropped images and that is the reason for their worse performance. They need to simultaneously localize the object as well as estimate the viewpoint. Can you please correct me if I am wrong?

YoungXIAO13 commented 4 years ago

Yes, I used GT boxes in all experiments as this paper focused on pose estimation on unseen objects providing the location in 2D image and its 3D cad model.

Meanwhile, other baselines on Pascal3D also used GT bounding boxes to crop the image, including StarMap.

For results on ObjectNet3D, we've explained that Xiang's result is estimated from the detected boxed in the caption of Table-1.

kartikgupta-at-anu commented 4 years ago

oh yes thanks for correcting me.