ilyasmg / sort-deepsort-yolov3-ros Goto Github PK

Hi @ilyasmg

I've noticed that since 0.47.0 numba only supports Python 3, and your track.py file uses Python 2 through the shebang #!/usr/bin/env python. Simple changing this to Python 3 does not seems to work as ROS Melodic doesn't support Python 3. When I do this I get the following error:

File "/home/ed-muthiah/catkin_ws/src/sort-deepsort-yolov3-ROS/sort_track/src/track.py", line 75
    rate = rospy.Rate(10)
                        ^
TabError: inconsistent use of tabs and spaces in indentation

Any suggestions to get this repo working for Python 3? Happy to help where I can :)

If there is particular version of numba, llvmlite, and other dependencies that still works please list them!

在darknet_ros下yolov3-tiny怎么和ros结合的，老哥有代码分享一下吗

Hi
Can I check is the deep sort result is based on real time video stream tracking (it does not come from a video file)? I am trying out the deep sort but kinda of confused as it seem to do the association based on predefined detection test file.

Thanks

usr/bin/env: “python”: 没有那个文件或目录

I have tried running the deepsort and it works great, however i realized that i wasnt using my GPU for TF. Hence i install TF gpu and had to lower the TF version from 2.7 to 2.2 for the gpu to work. However, by doing so, I was not getting any result printed after running the deepsort.

There are no error code and I have double checked the camera topic, the rqt graph also looks normal with the deepsort subscribing to the correct topic. However, no data was produce. Did anybody faced this issue before?

Any help would be greatly appreciated, thanks!

edits:
it seems to be an issue with the cudnn. I am running cuda 10.1 with cudnn 7.6.5, when i have delete the cudnn, the deepsort seems to be working fine, and when i reinstall cudnn, no data was produce. Hence i can narrow down that the issue is with the cudnn. However I am unsure what causes this issue.

Hello ilyas,

Thanks alot for your great work and effort. I knew that this package is no more supported, but i am using your code in my master thesis project and i am stuck in a point where i can not go ahead since 2 weeks. if you can help me i would be really thankful for you !

I got a error NameError: name 'detections' is not defined
I checked the track.py line 58, The detections was defined as global variable.
I did't kown what cause this issue.

[ERROR] [1662190947.981701]: bad callback: <function callback_image at 0xffff93023310>
Traceback (most recent call last):
  File "/opt/ros/noetic/lib/python3/dist-packages/rospy/topics.py", line 750, in _invoke_callback
    cb(msg)
  File "/home/nx/portable/catkin_workspace/src/sort-deepsort-yolov3-ROS-ros-noetic/sort_track/src/track.py", line 58, in callback_image
    cv2.rectangle(cv_rgb, (int(detections[0][0]), int(detections[0][1])), (int(detections[0][2]), int(detections[0][3])), (100, 255, 50), 1)
NameError: name 'detections' is not defined

The current code has some python2 dependency problems that are not easy to fix and not worth as python2 is deprecated.
I created a new branch ros-noetic with a docker environment to do the porting...currently I'm having some issues with building darknet_ros with opencv4.
If someone is able to build darknet_ros with opencv4 (then maybe it would be good to add the darknet_ros working as a submodule), next step would be to port track.py and track_deep.py to python3, saving dependencies in a requirements.txt so it will be easier to use in the future.
I added the sort and deep_sort packages as submodule in ros-noetic branch (they are both python3 compatible)

Hello, Thank you for your repository. I tried it using darknet_ros as mentioned in readme file, but on result I think, it labels every objects detected by darknet_ros as person even if it is not a person. Any suggestions what can I do for next step?

there is a error ,i use yolov3-tiny or yolov3, roslaunch sort_track sort.launch ,I can not solve it.

waiting for your reply,thanks.