Prediction takes too long

[Arunavameister]

Hello,
Firstly, thank you for making the source code of the suite open source.
I have a problem regarding the time taken for prediction. It is just above 2 seconds for a single image(on google colab with Telsa T4 gpu) which makes it almost impossible to use it for real time detection using an external camera.
Has anyone faced a similar issue?
Thanks

[yucheng-china]

the same to you, but i think that will refers to the vedio semantic segmentation, and we may need to use a visual tool like Rviz to display the real-time segmentation results, which is out of the functions of this repository