Tool/software:
Hi，
Now i use yolox_s_lite Model, When I train the model and deploy it to the target board，because the difference in picture background often leads to misjudgment，
Is there a good way to remove the image background during real-time object detection? thanks
e.g.
My training model is to detect black dots on the pen. Similar black dots on the background can also be detected. What I mean is whether there is a good way to remove the background or replace it with a uniform background before recognition so that it is not affected by the background. 
