Vision Encoder training traffic light state and stop sign state

[ni-eminen]

Hi!

I'm a bit confused about your way to get a target for the traffic light state. It seems you are using this clause:

    if len(measurements['is_red_light_present']) > 0:
        traffic_light_state = 0
    else:
        traffic_light_state = 1

but if multiple red lights are present, where the red lights don't concern the ego vehicle, won't this cause the target value to be wrong? I could imagine a scenario where there are three independent red lights in an intersection, and one of them will always be red. So then a red light is always present.

Am I interpreting the code wrong?

Cheers!

[deepcs233]

Hi!

The value measurements['is_red_light_present'] only includes red lights that affect the ego vehicle, not all lights in the intersection. So unrelated red lights won’t cause the target value to be wrong.