Does v12 work this way? Before E2E I might expect FSD to make explicit trajectory calculations, but does v12 work this way?
Also, regarding the processing of multiple camera feeds at the same. If the training is taken from human driving examples, and humans cannot see in all directions...