search

ethnhe / PVN3D

Code for "PVN3D: A Deep Point-wise 3D Keypoints Hough Voting Network for 6DoF Pose Estimation", CVPR 2020
MIT License
490 stars 105 forks source link

Own Dataset #18

Closed Michael187-ctrl closed 3 years ago

Michael187-ctrl commented 4 years ago

Hey Guys, great Code and Paper. I stumbled upon this a few weeks ago and got some really good results with the linemod classes. But now I want to try it for my own objects, can someone please explain the best way how to obtain the meshes , poses and so on for the training of my own dataset? Greetings Michael

ethnhe commented 3 years ago

You can generate some synthesis data with render scripts like pvnet_rendering and our raster triangle. For real data, you can read some dataset papers to see how they collect and label the data, like LineMOD, YCB-Video, tless, and so on. While it's not so easy to label the ground truth poses.