search

NVlabs / Deep_Object_Pose

Deep Object Pose Estimation (DOPE) – ROS inference (CoRL 2018)
Other
1.02k stars 287 forks source link

Training on Custom data & CPU machine #8

Open nikhiljaiswal opened 5 years ago

nikhiljaiswal commented 5 years ago

Hi, How can we train the network on our custom data? Is GPU required during test time also?

TontonTremblay commented 5 years ago

I will refer you to the newly uploaded training script: https://github.com/NVlabs/Deep_Object_Pose/blob/master/src/training/train.py.

SHZ2006 commented 5 years ago

Hi, Have you taken the pictures in UE by handling the camera by hand? Or there is a code to do it automatically? I mean when I put the object in UE and start the snapshots, just the environment changes and the object and camera are fixed. Shall I move the camera by hand for each picture manually? Best,

TontonTremblay commented 5 years ago

https://github.com/NVIDIA/Dataset_Synthesizer I will refer you to the NDDS tool for generating synthetic data. There is a component called the orbital camera, it will move the camera for you :P

avinashsen707 commented 4 years ago

I will refer you to the newly uploaded training script: https://github.com/NVlabs/Deep_Object_Pose/blob/master/src/training/train.py.

is link looks like error 404

mintar commented 4 years ago

The script location has changed, it's now here:

https://github.com/NVlabs/Deep_Object_Pose/blob/1ee8e0ddaa01886785ba176581582947008440d4/scripts/train.py

TontonTremblay commented 4 years ago

Are you having issues with NDDS or with dope itself. If you are having issues with NDDS, please make an issue on the NDDS github please.

avinashsen707 commented 4 years ago

@TontonTremblay let me look into it