liuyuan-pal / Gen6D

[ECCV2022] Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images
GNU General Public License v3.0
578 stars 73 forks source link

How to run the model with pictures of a LINEMOD "glue" object printed by a 3D printer #151

Open safsfsvvea opened 2 months ago

safsfsvvea commented 2 months ago

Thank you very much for your work! I want to inference the model with several photos of a "glue" object printed by a 3D printer, what should I do? If I simply replace data/LINEMOD/glue/JPEGImages to the directory of my own photos, there will be some errors, so I wonder what files under data/LINEMOD/glue do I need to change or define by myself? Thank you very much for your help!

safsfsvvea commented 2 months ago

I just modify both test.txt and train.txt to: LINEMOD/glue1/JPEGImages/000000.jpg LINEMOD/glue1/JPEGImages/000001.jpg LINEMOD/glue1/JPEGImages/000002.jpg and it seems that the detection results are very bad. ![Uploading image.png…]() I wonder whether do I need to produce a mask of my pictures by myself.