search

NVlabs / Deep_Object_Pose

Deep Object Pose Estimation (DOPE) – ROS inference (CoRL 2018)
Other
1.02k stars 286 forks source link

continue with a trained model #157

Open fehimkus opened 3 years ago

fehimkus commented 3 years ago

hello,

I wanted to ask 2 questions,

1- How can I continue training with a ready model, for example I have net_epoch_84, and I want to load this model. In python code I found this line to load but I can not be sure.

original 1308 if opt.net != '': 1309 net.load_state_dict(torch.load(opt.net))

changed 1308 if opt.net != '': 1309 pathh = "/content/drive/MyDrive/pthlar/net_epoch_84.pth" 1310 net.load_state_dict(torch.load(pathh))

2- I am exporting such a file from NDDS that contain only png type and you can see here. I did not use scene and approximately 5000 capture I stocked. Do you think it is proper way to train? I would be appreciated any help.

thank you

000000 000000 depth 16 000000 cs

TontonTremblay commented 3 years ago

This look good to me.

fehimkus commented 3 years ago

Why in this picture the object is green? I am using GroupActorManager_BP