Jetson Nano Implementation Issues

[frd-na]

Hi ahmetozlu!

Thank you for sharing your API on object counting using tensorflow.

I would like to use this for vehicle counting and pedestrian counting in a Jetson Nano. However, I am having issues implementing it.

My Jetson Nano is flashed with Jetpack 4.3 that contains OpenCV 4.1.1 , Python 3.6.9 and I have installed Tensorflow 2.0.0 based on these https://docs.nvidia.com/deeplearning/frameworks/install-tf-jetson-platform/index.html instructions from NVIDIA

I made some changes in the code of vehicle_counting.py to fix the compatibility of the tensorflow version. Then I ran the code but the the output video has some issues and the built-in player can't play it and says "cannot demultiplex stream". The output video is only 5.7 kB.

I think there are some issues in writing the video. Is it compatibility issues with OpenCV version? What is the OpenCV and Tensorflow versions you used when you implemented this?

Any help is greatly appreciated.