ZeroDivision error about yolov8_ros HOT 7 CLOSED

That's a bit strange. Can you check the image messages published in the image topic with ros2 echo?

from yolov8_ros.

It looks like this:

header:
stamp:
sec: 0
nanosec: 0
frame_id: ''
height: 320
width: 320
encoding: bgr8
is_bigendian: 0
step: 0
data:

• 103
• 120
• 140
• 255
• 103
• 120
• there is a lot of similar lines with int8
• 255
• '...'

from yolov8_ros.

I edited source code of your package by edit some parameters in model.predict: augment=False and imgsz=(320,320) so it wont transform images, but it still do

from yolov8_ros.

You should fill all the fields of the messages, I mean the frame_id, the stamp and the step.

from yolov8_ros.

Frame_id and stamp is not fillable according to ros docs
I tried to fill step, but I dont actually get what it needs
In docs it described as "Full row length in bytes", and I tried to use len() or sizeof() functions on image (that I get by self.__cam.getImage()), but its not working
[yolov8_node-1] File "/opt/ros/humble/local/lib/python3.10/dist-packages/cv_bridge/core.py", line 180, in imgmsg_to_cv2 [yolov8_node-1] im = np.ndarray(shape=(img_msg.height, int(img_msg.step/dtype.itemsize/n_channels), n_channels), [yolov8_node-1] TypeError: buffer is too small for requested array

from yolov8_ros.

Why are they not fillable? As you are publishing your own image messages, you should fill it with the acquisition time and the optical frame of your simulated camera. This is important for the 3D version, which uses message synchronization and the tf.

Anyway, the step is the full row length in bytes, that is the Image Width * Bytes per Pixel. In your case, the width is 320 and the bytes per pixel are obtained from the encoding type, that is BGR8 (3 bytes per pixel). So your step is 960 bytes.

from yolov8_ros.

Thanks a lot!
It works fine for me with step, thanks for explanation, I got that stamp and id should be filled via std_msgs/Header, but it worked without it, so I decided to continue my work without it
It has issues with objects I detect, all of them has the same class_name, its strange because I tested model separately and it worked fine. But its not about issue I started, thanks a lot for your help!

from yolov8_ros.