search

NVlabs / Deep_Object_Pose

Deep Object Pose Estimation (DOPE) – ROS inference (CoRL 2018)
Other
1.01k stars 283 forks source link

cannot get any image from camera #7

Open qq4060 opened 5 years ago

qq4060 commented 5 years ago

hi I just run and cannot get any image from camera ,ut shows:

~/catkin_ws$ rosrun dope camera.py 
Publishing images from camera 0 to topic '/dope/webcam_rgb_raw'...
Ctrl-C to stop
^Cqq4060@qq4060-To-be-filled-by-O-E-M:~/catkin_ws$ rosrun dope camera.py 
Publishing images from camera 0 to topic '/dope/webcam_rgb_raw'...
Ctrl-C to stop

and also shows 

~/catkin_ws$ rosrun dope dope.py
Loading DOPE parameters from '/home/qq4060/catkin_ws/src/dope/config/config_pose.yaml'...
    Parameters loaded.
Loading DOPE model '/home/qq4060/catkin_ws/src/dope/weights/soup_60.pth'...
    Model loaded in 4.77608513832 seconds.
Running DOPE...  (Listening to camera topic: '/dope/webcam_rgb_raw')
Ctrl-C to stop

Thank you very much!

TontonTremblay commented 5 years ago

This looks great to me. Do you see the /dope/webcam_rgb_raw topic in rviz?

qq4060 commented 5 years ago

OH I CANNOT SEE ANY OTHER THINGS OPEN OR RVIZ

qq4060 commented 5 years ago

hi it is like this?Thanks image

yuzw213 commented 5 years ago

Hi, have you tried to train your model with pytorch?

qq4060 commented 5 years ago

Hi, have you tried to train your model with pytorch?

SORRY HOW TO TRAIN? use this line"$ rosrun dope dope.py [my_config.yaml] # Config file is optional; default is config_pose.yaml"?

yuzw213 commented 5 years ago

Well. Can you see the effect of dope in rviz? At present, I can only use six models provided by invidia, but I want to train my own model, which needs pytorch

qq4060 commented 5 years ago

CAN YOU SHOW ME YOUR EFFECT ? I CANNOT SEE ANYTHING IN MY RVIZ.

it said

This is the official DOPE ROS package for detection and 6-DoF pose estimation of known objects from an RGB camera. The network has been trained on the following YCB objects: cracker box, sugar box, tomato soup can, mustard bottle, potted meat can, and gelatin box. For more details, see our CoRL 2018 paper and video.

Note: Currently this package contains inference only.

yuzw213 commented 5 years ago

First you may need the six models or pictures provided by invidia. In rviz, set the image topic to: / dope / rgb_points, and then point the camera at your model or picture.

qq4060 commented 5 years ago

Thanks, Is the six models or pictures provided by invidia in this step ? YCB 3D Models DOPE returns the poses of the objects in the camera coordinate frame. DOPE uses the aligned YCB models, which can be obtained using NVDU (see the nvdu_ycb command).

yuzw213 commented 5 years ago

Yes, but you can search the pictures of the six models online first. Nvdu and NDDs are used to train YCB model data. It seems that pytorch will be used in the end. It's a bit troublesome. We can exchange and study together at that time. I hope it will be helpful to you.

qq4060 commented 5 years ago

how about this they said no train code

https://github.com/NVlabs/Deep_Object_Pose/issues/5

yuzw213 commented 5 years ago

No, you can study it. I'm also trying to train my model.

qq4060 commented 5 years ago

OK Thanks a lot!