one or more processors; and

one or more non-transitory computer-readable media storing instructions that, when executed by the one or more processors, cause the system to perform operations comprising:

receiving a first frame of image data that was generated, at a first instance of time, by an image sensor associated with a first vehicle, the first frame representing a second vehicle in an environment in which the first vehicle is operating at the first instance of time;

receiving a second frame of image data that was generated by the image sensor at a second instance of time after the first instance of time, the second frame representing the second vehicle in the environment at the second instance of time;

inputting the first frame and the second frame into a machine-learned model that is configured to determine a state of one or more doors of the second vehicle;

receiving, from the machine-learned model, an output indicating the state of the one or more doors of the second vehicle;

determining, as an association, that a pedestrian is associated with the second vehicle;

determining, based at least in part on the state of the one or more doors of the second vehicle and the association of the pedestrian with the second vehicle, a predicted trajectory associated with the pedestrian, wherein the predicted trajectory comprises a predicted pedestrian destination proximate a door of the second vehicle; and

controlling the first vehicle based at least in part on the output and the predicted trajectory, wherein controlling the first vehicle comprises altering a trajectory associated with the first vehicle based at least in part on avoiding the pedestrian.