search

DC1991 / G2L_Net

MIT License
118 stars 13 forks source link

About making 'best_obj.pt' on my own #32

Open XLXIAOLONG opened 2 years ago

XLXIAOLONG commented 2 years ago

Hi, @DC1991 . Thanks for your outstanding work and sharing! I have trained the 'cat'(models 6 in this project), but I don't know how to make best_6.pt by myself. Can you give me some suggestion? Any word would be helpful!

Best wishes!

XLXIAOLONG commented 2 years ago

I have trained obj_6 by yolov3 and G2L_NET, but the test code(test_linemod.py) doesn't work.

The error is:
RuntimeError: Error(s) in loading state_dict for Darknet: Missing key(s) in state_dict: "module_list.0.conv_0.weight", xxxxxxxxxxxxxxxxxxxxx Unexpected key(s) in state_dict: "total_ops", "total_params",xxxxxxxxxxxxxxxxxxxx

Mahmoud-Ali-FCIS commented 1 year ago

@XLXIAOLONG @DC1991 How you trained G2L?

I tried to run trained_G2l.py but i had this problem: RuntimeError: Given groups=1, weight of size 64 3 1 1, expected input[64, 6, 1, 1500] to have 3 channels, but got 6 channels instead

Notes that I followed all the instructions and i download the training data and put it in the right folder.

XLXIAOLONG commented 1 year ago

@XLXIAOLONG @DC1991 How you trained G2L?

I tried to run trained_G2l.py but i had this problem: RuntimeError: Given groups=1, weight of size 64 3 1 1, expected input[64, 6, 1, 1500] to have 3 channels, but got 6 channels instead

Notes that I followed all the instructions and i download the training data and put it in the right folder.

Sorry for the late reply. Because some reasons, I haven't train G2L on my own dataset. But i had the same problem. The reason seem to be the point cloud. The 6 channels contain color information, but G2L need coordinate only. Extract 3 channels position from point cloud may solve the problem.

Best wishes! xiaolong