While analyzing the prediction code, when traffic red light is available car stops/traffic sign board it recognizes.

How traffic light/traffic signs are getting detected since only sequence images and directions are used to predict the affordances? It is not using measurements data for prediction..

Is this is a fasle proof solution......?