Open HappySamuel opened 1 year ago
There are several modalities of obstacle detection. Apart from rmf_human_detector_oakd
, the other packages are hardware agnostic and perform inference based on ROS 2 sensor msg data. The realsense-ros driver can help publish RGB images which you can pass to rmf_human_detector
. You may even be able to convert the depth pt cloud into a laser scan msg and pass that data to rmf_obstacle_detector_laserscan
.
If you're willing to contribute code, I can also guide you on enhancing rmf_human_detector
to
a) make it a more generalized detector (not just humans)
b) Incorporate (optionally) a depth stream to obtain 3D pose data of the inferences. Right now some heuristics are used to estimate depth from 2D RGB image.
Hi @Yadunund
That would be great to have rmf_human_detector to be a) more generalized detector b) incorporate with depth stream
I am willing to contribute the code under your guidance on achieving this.
Best, Samuel
Hi @HappySamuel sorry for the overdue response. I was occupied with other efforts. I should have more time for this towards the end of this month. Will put a plan together. Feel free to ping here again if you don't hear from me.
Hi @Yadunund Good to hear from you again. Looking forward to work with you together on this.
Best, Samuel
@HappySamuel sorry for the long silence here.
Open-RMF is part of GSoC 2024 and one of the projects is to cleanup and improve the obstacle detection pipeline. You may want to checkout https://discourse.ros.org/t/attention-students-open-robotics-google-summer-of-code-2024-projects/
I've also opened #22 with a list of items to work on. The goal will be to support a wide range of sensors and detection algorithms via a plugin based system. Let me know if you're still keen on contributing.
Hi
Can this rmf_obstacle_detectors work with any Realsense D400 series Cameras? If i wanna use D400 series cameras, what are the steps to integrate them up?
Best, Samuel