[BUG] Image is wrongly formed: step < width * byte_depth * num_channels or 2560 != 1280 * 1 * 3

[BryanBetancur]

Describe the bug Hello, I'm testing a OAK-D-PRO-W-POE in order to check if is worth it to change our Realsense robots cameras to OAK cameras but I'm getting this error when I set i_low_bandwidth: true, if it's in false the camera works properly but the fps are so low.

[oak_container-1] [INFO] [1741632856.966561767] [oak]: Starting camera.
[oak_container-1] [INFO] [1741632856.981630463] [oak]: No ip/mxid specified, connecting to the next available device.
[oak_container-1] [INFO] [1741632867.529489182] [oak]: Camera with MXID: 14442C103110F4D600 and Name: 172.16.40.120 connected!
[INFO] [1741632867.529634144] [oak]: PoE camera detected. Consider enabling low bandwidth for specific image topics (see Readme->DepthAI ROS Driver->Specific camera configurations).
[oak_container-1] [INFO] [1741632867.592237216] [oak]: Device type: OAK-D-PRO-W-POE
[oak_container-1] [INFO] [1741632867.623092985] [oak]: Pipeline type: RGBD
[oak_container-1] [INFO] [1741632868.984961750] [oak]: Finished setting up pipeline.
[oak_container-1] [INFO] [1741632869.399505025] [FFMPEGEncoder]: enc: libx264 prof:  preset: 
[INFO] [1741632869.399587362] [FFMPEGEncoder]: qmax: 10 bitrate: 8242880 gop: 15
[oak_container-1] [INFO] [1741632869.796394895] [FFMPEGEncoder]: enc: libx264 prof:  preset: 
[INFO] [1741632869.796475057] [FFMPEGEncoder]: qmax: 10 bitrate: 8242880 gop: 15
[oak_container-1] [INFO] [1741632871.160678255] [oak]: Camera ready!
[oak_container-1] terminate called after throwing an instance of 'cv_bridge::Exception'
  what():  Image is wrongly formed: step < width * byte_depth * num_channels  or  2560 != 1280 * 1 * 3

Minimal Reproducible Example I'm using the example launch camera.launch.py with the following params file:

/oak:
  ros__parameters:
    camera:
      i_nn_type: none
    rgb:
      i_low_bandwidth: true
 

Expected behavior After run ros2 launch depthai_ros_driver camera.launch.py it will log: what(): Image is wrongly formed: step < width * byte_depth * num_channels or 2560 != 1280 * 1 * 3.

Pipeline Graph

14442C103110F4D600_pipeline.json

System specifications

• Jetson Orin AGX 64GB
• OAK-D-PRO-W-POE
• depthai-ros: v2.11.0-humble

[Serafadam]

Thank you for the report, investigating this now.

[Serafadam]

Fix: https://github.com/luxonis/depthai-ros/pull/689

[HJGrant]

I have a similar problem, when I set i_low_bandwith: true, I get the following error when I run camera.launch.py:

[depthai] [error] Callback with id: 1 throwed an exception: Converted type not supported!

If I set i_low_bandwith: false, I don't get the error but the fps are really low, lik 12 fps, when I expect 30 fps.

when I set i_low_bandwith: true for the stereo node, I get the following warning:

'[VideoEncoder(13)] [warning] Arrived frame type (14) is not either NV12 or YUV400p (8-bit Gray)'

and nothing is published to the /oak/stereo/raw_image and /oak/stereo/camera_info topics. Setting i_low_bandwith: false for both nodes solves these issues, but the FPS are very slow.

I am using the OAK-D PoE.

[Serafadam]

@HJGrant do you also get the same results when running from this branch?

[HJGrant]

Unfortunately I get the same error:

git branch
  humble
* humble_color_order

install/my_oakd_launch/share/my_oakd_launch/config/rgbd.yaml

    rgb:
      i_publish_topic: true
      i_resolution: 1080P
      i_fps: 25.0
      i_low_bandwidth: true
      i_output_isp: true
      i_isp_den: 3
      i_isp_num: 2
    stereo:
      i_publish_topic: true
      i_low_bandwidth: true
      i_align_depth: true
      i_lr_check: true
      i_subpixel: true

leads to:

[depthai] [error] Callback with id: 1 throwed an exception: Converted type not supported!
[VideoEncoder(14)] [warning] Arrived frame type (14) is not either NV12 or YUV400p (8-bit Gray)

[Serafadam]

Hi, unfortunately subpixel mode (which produces 16Bit RAW images (type 14)) doesn't currently work with low_bandwidth mode as VideoEncoder that we use in the firmware doesn't support it. Though it should be turned off automatically when enabling low_bandwidth mode. In the logs you should see something like [component_container-1] [INFO] [1741864190.175553025] [oak]: Subpixel disabled due to low bandwidth mode Did you make changes in the source code? Could you clean up and rebuild your workspace so that we are sure older version of code isn't used?

[HJGrant]

After cleaning and rebuilding my workspace I can confirm that this branch fixes the problem, thank you!

[dwffls]

I got the same problem here, however it is when using the i_publish_left_rect or i_publish_right_rect parameter. With this the step appears to be set to 0, and thus the error becomes: step < width * byte_depth * num_channels or 0 != 1280 * 1 * 1

When publishing a left image with left.i_publish_topic set to true the step is correctly set to 1280;

I've been looking in the code on where this step is set but can't find it anywhere. @Serafadam could you help me out here?

[Serafadam]

Hi, sorry for the delay, I was on vacation, could you share your parameters file?

[dwffls]

Not a problem, hope you had a nice vacation!

This is the parameter file we currently use

camera:
  i_nn_type: none
  i_pipeline_type: Depth
  i_publish_tf_from_calibration: true
  i_enable_ir: true

pipeline_gen:
  i_enable_sync: true

rgb:
  i_synced: true 
  i_width: 1280

  r_set_chroma_denoise: true
  r_chroma_denoise: 2 # denoising strength (0–3)
  r_set_luma_denoise: true
  r_luma_denoise: 2
  i_publish_topic: true
  r_set_man_exposure: false
  r_set_man_focus: false
  r_set_man_whitebalance: false
  r_whitebalance: 3300



stereo:
  i_extended_disp: true
  i_subpixel: true
  i_synced: true
  i_publish_compressed: False
  i_align_depth: true
  
  # Ensure depth topic is published
  i_publish_topic: true
  i_output_disparity: true
  
  # Enable and configure depth filtering for better outdoor performance
  i_enable_spatial_filter: true
  i_enable_temporal_filter: true
  i_enable_threshold_filter: true
  i_enable_brightness_filter: true #false
  
  # Increase depth range for parking lot distances
  i_depth_preset: HIGH_ACCURACY
  i_stereo_conf_threshold: 100   # Lower = more depth points (was 150)
  i_disparity_width: DISPARITY_96  # Changed from DISPARITY_128 to supported value
  
  # Configure depth filtering
  i_depth_filter_size: 5
  i_bilateral_sigma: 0
  
  # Add depth range limits for parking lot
  i_depth_min_m: 0.3    # Minimum depth in meters
  i_depth_max_m: 25.0   # Maximum depth in meters
  
  # Left-right consistency check
  i_lr_check: true
  i_lrc_threshold: 5
  
  # Ensure proper rectified images are published
  i_publish_left_rect: true
  i_publish_right_rect: true
  i_publish_synced_rect_pair: true