I'm into a project wherein i have to detect the object using using yolov5 and put that object inside the map .Can anyone here please provide me a pathway to do so,It will be very useful.
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
7
Here is my idea :
- Take the depth image provided by your camera
- If you only want to detect objects (and not other obstacles), modify this depth image by setting pixels outside the objects bounding boxes to a large value.
- use depthimage_to_laserscan to convert this depth image to a laserscan.
- Integrate the laserscan with your costmap
If performance is not an issue, you can directly integrate the depth image as a Voxel Layer by converting it to a PointCloud (http://wiki.ros.org/depth_image_proc) so that you have 3D object avoidance (meaning that your device will be able to say "ok there is an obstacle but it is above me so I can still go through"). If you don't need this feature, convert it using step 3.
You can also take a look at http://wiki.ros.org/spatio_temporal_voxel_layer for the integration of your laser scan or depth image with your costmap.
-
$\begingroup$ the objects which i detect using yolo need to be put as an obstacle inside the global costmap and later then use than map for navitgation purpose $\endgroup$ Nov 10 at 10:20
-
$\begingroup$ Do you use a stereo camera in order to get distance between you and the object ? $\endgroup$ Nov 10 at 10:26
-
$\begingroup$ i'm using intelsense 435d to serve the purpose $\endgroup$ Nov 10 at 10:27
-
$\begingroup$ thanks for the support really appreciate it..I need a little detail on the process between point 2 and 3 $\endgroup$ Nov 11 at 2:40
-
$\begingroup$ Imagine your object detection detects humans only and there is a dog in your image aswell (you don't detect it using yolo). Do you want to consider this dog as an obstacle aswell or not ? If yes, don't modify the depth image. If not, modify it by setting everything outside your bounding box (of the human) to a large value so that it won't be considered as an obstacle. $\endgroup$ Nov 11 at 9:56
$\begingroup$
$\endgroup$
1
As an example of prior art:
Here's a guy that did it with Tensor flow and a kinect camera for the object and depth detection. YOLO and Realsense is not so different in application.
https://github.com/douglasrizzo/dodo_detector_ros
I'm not saying his code will work for you, but always a good idea to study those that came before you.
