LSD-Slam capture does not start

[holzkohlengrill]

When I am trying to start LSD-Slam like:

rosrun lsd_slam_core live_slam image:=/image_raw _calib:='<path_to_my_calibrationfile.cfg'

I get the following output (no window appears, ...):

Reading Calibration from file /home/marcel/rosbuild_ws/package_dir/lsd_slam/lsd_slam_core/calib/OpenCV_example_calib.cfg ... found! found OpenCV camera model, building rectifier. Input resolution: 752 480 In: 500.000000 500.000000 350.000000 240.000000 0.000000 0.000000 0.000000 0.000000 Out: Crop Output resolution: 640 480 Started mapping thread! Started constraint search thread! Started optimization thread

Another point (which could be related to this issue) is that when I view a pointcloud dataset I got colour information (just the viewer is shown).

Thanks.

[kilmarnock]

The window appears as soon as you provide images on the right topic

[xiaokeme]

@kilmarnock ,I encountered the same question as @holzkohlengrill ,but how to provide images on the right topic?

[xiaokeme]

the steps I did is shown: https://github.com/tum-vision/lsd_slam/issues/141

[kilmarnock]

@xiaokeme , try to set up your camera without using lsd_slam. Verify your setup:
rostopic hz /image

should give you the current frames per second on this topic. If you do not get any image, try to ask your question in ros-answers, it is not lsd_slam related then but rather a problem of your camera setup.

Your steps in 141 show: [uvc_camera-1] process has died [pid 4657, exit code -11, cmd Dead camera -> no images. Try to setup without lsd_slam.

[xiaokeme]

@kilmarnock ,thank you very much!

[mihaibujanca]

@kilmarnock I'm having the same issue. The topic I am using is clearly published, since I can view it using image_view and I can do

rostopic hz /image_raw

For me it only gets as far as

found ATAN camera model, building rectifier.
Input resolution: 1920 1080
In: 0.527334 0.827306 0.473568 0.499436 0.000000
NO RECTIFICATION
Output resolution: 752 480
Prepped Warp matrices
Started mapping thread!
Started  constraint search thread!
Started optimization thread 

[kilmarnock]

@mihaibujanca lsd_slam waits for its images at the "image" topic. You can either remap it:

<remap from="image" to="/image_raw" /> and check it with rostopic hz /image_raw

or provide your image at /image and check it with rostopic hz /image

I corrected my post above, sorry for the mistake.

[mihaibujanca]

@kilmarnock I am providing lsd slam with the image topic when I run it:

rosrun lsd_slam_core live_slam image:=/image_raw _calib:=mycalib.cfg

So I don't think that's the problem

[kilmarnock]

what is the output of rostopic info /image_raw?

[mihaibujanca]

@kilmarnock

rostopic info /image_raw
Type: sensor_msgs/Image

Publishers:
* /video2ros (http://mihai-Aspire-E1-571G:41184)

Subscribers:

* /image_view_1444123124434232 (http://mihai-Aspire-E1-571G:60608)

Video2ros (package rbx1/rbx1_vision) is just publishing a video I recorded with a pi camera - at some point I'll try to stream the video from a pi and have it processed on my laptop. Looks like lsd_slam is not subscribing to it, though image_view has no problem with it, the image is displayed properly

[kilmarnock]

Your assumption is right: lsd_slam is not subscribing. Try to use a launch file and remap the image topic in there:

 <node pkg="lsd_slam_core" type="live_slam" name="lsd_slam" output="screen">
        <remap from="image" to="/image_raw"/>
        <remap from="camera_info" to="/mycam/camera_info"/>
        <param name="calib" value="/path/to/config.cfg"/>
  </node>

You have to end providing images like so:

rostopic info /image_raw
Type: sensor_msgs/Image

Publishers: 
 * /somenode (http://somehost:56671/)

Subscribers: 
 * /lsd_slam (http://somehost:57934/)
 * /rviz_1444924448230228527 (http://otherhost:55785/)

[mihaibujanca]

@kilmarnock Thanks, indeed remapping instead of passing it as a param worked.

All I have to do now is get a good calibration file, since I'm getting TRACKING LOST for frame 29 (0.09% good Points, which is 41.50% of available points, DIVERGED)! with all videos.

Thanks a lot .