Question about training my own dataset

XuyangBai / D3Feat

[TensorFlow] Official implementation of CVPR'20 oral paper - D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features https://arxiv.org/abs/2003.03164

MIT License

261 stars 38 forks source link

Question about training my own dataset #21

Closed GitHubOKay closed 3 years ago

GitHubOKay commented 4 years ago

Hi. I have one question about how to train my own dataset. I see the python code 'datasets/cal_overlap.py' you provided, and I am confused about the 'scenelist{split}.txt' and the '{ind}.pose.npy' files. It is will be helpful for me to understand, if you provides these files to me. Thanks very much. Have a nice day.

XuyangBai commented 4 years ago

Hi @GitHubOKay 3DMatch provided 54 scenes including 8 scenes for testing, the scene_list_{split}.txt indicates which scene is for training or testing, I derive it from split.txt from the 3DMatch website. And the pose.npy is the ground truth pose of fused point cloud fragments generated from the TSDF fusion step, you can find the code in my another repo

Best, Xuyang.

GitHubOKay commented 4 years ago

Hi, @XuyangBai. Thanks for your reply. I will try.