Open 512792354 opened 2 years ago
Hi, you may check the intermediate results according to https://github.com/liuyuan-pal/Gen6D/issues/4#issuecomment-1186136894 The common issue causing incorrect results is the incorrect cropping of the object point cloud.
Hi, you may check the intermediate results according to #4 (comment) The common issue causing incorrect results is the incorrect cropping of the object point cloud.
Thanks for suggestion,I'll try with that.
Hi,I followed the instructions under "Pose estimation on custom objects", and tried the code in Ubuntu 18.04. I tried with the mouse image. But when I ran predict.py, it turned out with a wrong result that the estimated pose shaking around the mouse, instead of focused on the mouse. I have tried download the mouse_processed folder and run it directory. Also,I tried to manually split reference video with colmap and crop with cloudcompare. But both of them finally got the same result. Could you help with that? I used python3.9 in anaconda,pytorch is '1.12.1+cu102' version.