can't see bounding boxes

[Habibjackson]

I couldn't see any bounding box can anyone please suggest me a solution and also the info flag is not working.

[jamshaidsohail5]

Hi. I am facing the same issue on my local PC with GeForce RTX 3060 (6GB) but when I run the code on Goolge Colab, everything seems to work very much fine. If you are able to dig into the issue and find a solution please do share.

[jamshaidsohail5]

Hey I resolved the issue by upgrading the tensorflow version to 2.5.

pip install tensorflow-gpu==2.5

The above command did the job for me. Good luck.

[Habibjackson]

Thanks for your suggestion brother but i am already having tensorflow-gpu 2.8.0 still no bounding boxes.

[alantseone]

I also couldn't see any bounding box and I print the pred_bbox, only the first frame have output pred_bbox. I had this problem before starting. QObject::moveToThread: Current thread (0x563bcd421950) is not the object's thread (0x563b8b6d2fb0). Cannot move to target thread (0x563bcd421950)

I don't know it is affects the result?

[abdelmalek0]

tensorflow-gpu 2.8.0 yolov4-tiny I couldn't see any bounding boxes

[abdelmalek0]

tensorflow-gpu 2.8.0 yolov4-tiny I couldn't see any bounding boxes

I used tensorflow-gpu 2.3.0 and It worked finally

[Wetu-Vexo]

Hello guys, I also face the same issue with the gpu, but the code work fine on the cpu. So i try to convert the model with tensorflow cpu and run the object_tracking with tensorflow gpu.

[deshwalmahesh]

Yes, it is working fine with CPU. I there is an issue with the working of GPU tensorflow due to dependency

[kevin5k]

Hello guys, I also face the same issue with the gpu, but the code work fine on the cpu. So i try to convert the model with tensorflow cpu and run the object_tracking with tensorflow gpu.

I confirm this method works and I managed to get the bounding box now!!! Specifically, I performed the save_model.py command using tf-cpu (i.e., tensorflow==2.3.0) environment. For real-time inference, I use the tf-gpu (i.e., tensorflow==2.4.1) environment. Note that the libraries are created via anaconda environment only (i.e., no pip required).

Take the demo.avi as an example, the mean real-time inference performance I observed are 2fps (tf-cpu) as opposed to 22fps (tf-gpu) on a 2080Ti.

Hope this information helps everyone else.

[kevin5k]

tensorflow-gpu 2.8.0 yolov4-tiny I couldn't see any bounding boxes

I used tensorflow-gpu 2.3.0 and It worked finally

For the benefit of people like me, note the OS dependency differences for tensorflow-gpu package, I observed the following based on conda-forge channel:

• (Windows OS) tensorflow-gpu -> {2.3.0, 2.5.0, 2.6.0} are available
• (Linux OS - Ubuntu) tensorflow-gpu -> {2.2.0, 2.4.1, 2.6, 2.7, 2.8} are available

[JDmoric]

Hello guys, I also face the same issue with the gpu, but the code work fine on the cpu. So i try to convert the model with tensorflow cpu and run the object_tracking with tensorflow gpu.

I confirm this method works and I managed to get the bounding box now!!! Specifically, I performed the save_model.py command using tf-cpu (i.e., tensorflow==2.3.0) environment. For real-time inference, I use the tf-gpu (i.e., tensorflow==2.4.1) environment. Note that the libraries are created via anaconda environment only (i.e., no pip required).

Take the demo.avi as an example, the mean real-time inference performance I observed are 2fps (tf-cpu) as opposed to 22fps (tf-gpu) on a 2080Ti.

Hope this information helps everyone else.

Hello, I did as your tip that performed the save_model.py command using tf-cpu ( tensorflow==2.3.0) environment. For real-time inference, I use the tf-gpu (tensorflow-gpu==2.3.0) environment. But there is no detection.

How to deal with that.

Best wishes

[adityachintala]

Hi. I am facing the same issue on my local PC with GeForce RTX 3060 (6GB) but when I run the code on Goolge Colab, everything seems to work very much fine. If you are able to dig into the issue and find a solution please do share.

Could you once state the tensorflow-gpu and opencv-python versions you were using back then to run the deepSORT model?

[adityachintala]

Hello guys, I also face the same issue with the gpu, but the code work fine on the cpu. So i try to convert the model with tensorflow cpu and run the object_tracking with tensorflow gpu.

I confirm this method works and I managed to get the bounding box now!!! Specifically, I performed the save_model.py command using tf-cpu (i.e., tensorflow==2.3.0) environment. For real-time inference, I use the tf-gpu (i.e., tensorflow==2.4.1) environment. Note that the libraries are created via anaconda environment only (i.e., no pip required). Take the demo.avi as an example, the mean real-time inference performance I observed are 2fps (tf-cpu) as opposed to 22fps (tf-gpu) on a 2080Ti. Hope this information helps everyone else.

Hello, I did as your tip that performed the save_model.py command using tf-cpu ( tensorflow==2.3.0) environment. For real-time inference, I use the tf-gpu (tensorflow-gpu==2.3.0) environment. But there is no detection.

How to deal with that.

Best wishes

Did you find the solution to this? Even I'm facing the same issues.

[adityachintala]

tensorflow-gpu 2.8.0 yolov4-tiny I couldn't see any bounding boxes

I used tensorflow-gpu 2.3.0 and It worked finally

Opencv_python version?