Hi,
I read the deepstream_test2.py in deepstream’s python example code. It’s a detector + classifier tasks but I found something strange to me.
The problem is about elements’ link between tracker and classifier(car’s color, car’s type).
In Code:
tracker–>classifier1–>classifier2–>classifier3
But I think the dataFlow should be:
tracker–>classifier1
tracker–>classifier2
tracker–>classifier3
classifier1–>nvvidconv
classifier2–>nvvidconv
classifier3–>nvvidconv
How do I know or control the classifier’s output? I think the classifier’s output should be probability of classes. As a result, the classifier 2 can’t get any image pixel information from classifier1 and nvvidconv can’t get all the predict results from classifier1 + classifier2 + classifier3. Nvvidconv only received the predict result from classifier3.