fail to use function in cv_bridge

[wyixiang]

When I use some function in cv_bridge, I receive an importerror:

Traceback (most recent call last):
File "/home/yixiangw/gym-gazebo/gym_gazebo/envs/turtlebot/gazebo_circuit2c_turtlebot_camera_nn.py", line 217, in _reset
cv_image = CvBridge().imgmsg_to_cv2(image_data, "bgr8")
File "/home/yixiangw/gym-gazebo/gym_gazebo/envs/installation/catkin_ws/src/vision_opencv/cv_bridge/python/cv_bridge/core.py", line 163, in imgmsg_to_cv2
dtype, n_channels = self.encoding_to_dtype_with_channels(img_msg.encoding)
File "/home/yixiangw/gym-gazebo/gym_gazebo/envs/installation/catkin_ws/src/vision_opencv/cv_bridge/python/cv_bridge/core.py", line 99, in encoding_to_dtype_with_channels
return self.cvtype2_to_dtype_with_channels(self.encoding_to_cvtype2(encoding))
File "/home/yixiangw/gym-gazebo/gym_gazebo/envs/installation/catkin_ws/src/vision_opencv/cv_bridge/python/cv_bridge/core.py", line 91, in encoding_to_cvtype2
from cv_bridge.boost.cv_bridge_boost import getCvType
ImportError: dynamic module does not define module export function (PyInit_cv_bridge_boost)

so how can I solve it ?

[walid19]

i am having the same problem did you find the solution?

[wyixiang]

@walid19 In my try, cv_bridge is available for python2 and cannot be used in py3. You can try running in python2.

[Akashbaskaran]

Was anyone able to use python3 and solve this issue?

[ghost]

Found a way around. This helped http://www.programmersought.com/article/4908420840/

[qgallouedec]

You could use numpy directly, instead of cv_bridge

import numpy as np
im = np.frombuffer(image_data.data, dtype=np.uint8).reshape(image_data.height, image_data.width, -1)

[felixvd]

↑↑↑ A real MVP, just like in the legends

If you're missing the desired_encoding parameter, use cv2.cvtColor with these codes.

import cv2
im_gray = cv2.cvtColor(im, cv2.COLOR_RGB2GRAY)

[nerdneilsfield]

Is there anyway to implement the same thing as CvBridge().cv2_to_imgmsg rather than using cv_bridge?

[andrewjong]

A different solution worked for me. I'm also trying to use Python 3 with CVBridge. I'm using conda as my environment manager. For me the solution was to configure catkin like so:

catkin config \
-DPYTHON_EXECUTABLE=/home/andrew/.miniconda3/envs/myenv/bin/python \
-DPYTHON_INCLUDE_DIR=/home/andrew/.miniconda3/envs/myenv/include/python3.7m/ \
-DPYTHON_LIBRARY=/home/andrew/.miniconda3/envs/myenv/lib/libpython3.7m.so

After adding these CMake Args, I no longer get the error. Note if you're using a different version of conda / conda-python, you may have to adjust your -DPYTHON_INCLUDE_DIR and -DPYTHON_LIBRARY to point to the correct files as they're slightly different in some versions. I'm using ROS melodic.

[kilichzf]

I was having the same error message but not with this repository.
I am writing this because it may help somebody.
In my case I forgot to source cv2_bridge package for python 3 and trying to use python 2 one.

[tamerlan-b]

Beautiful solution is to use ros_numpy. It is similar to the function of cv_bridge, but without the dependency on cv2:

import ros_numpy
np_img = ros_numpy.numpify(ros_img_msg)

[arthurBricq]

If you are trying to not use cv_bridge when publishing an image to ROS (the other direction as the one of this issue): ndarray (or cv2) --> ROS, here is a snippet of code that does the job using the methods mentionned just above.

Assuming that img is a numpy array containing your image:

ros_image = Image(encoding="mono8")
# Create the header
ros_image.header.stamp = rospy.Time.from_sec(t)
ros_image.header.frame_id = id
# Fill the image data 
ros_image.height, ros_image.width = img.shape
ros_image.data = img.ravel().tobytes() # or .tostring()
ros_image.step=ros_image.width

[DanielTakeshi]

@qgallouedec where are you getting image_data in your example code? I have some "msg" which I am using with bridge.imgmsg_to_cv2(msg, 'bgr8') where bridge = CvBridge().