def parse_traffic_lights(self, traffic_lights):
# Each entry of traffic lights is a dictionary that contains four
# items, the id, state, position, and trigger volume of the traffic
# light.
# WARNING: Some of the methods in the TrafficLight class may not work
# here (e.g. methods that depend knowing the town we are in).
traffic_lights_list = []
traffic_light_labels = {
0: TrafficLightColor.RED,
1: TrafficLightColor.YELLOW,
2: TrafficLightColor.GREEN
}
for traffic_light_dict in traffic_lights.values():
traffic_light_id = traffic_light_dict['id']
traffic_light_state = traffic_light_labels[
traffic_light_dict['state']]
# Trigger volume is currently unused.
# traffic_light_trigger_volume = \
# traffic_light_dict['trigger_volume']
location = pylot.utils.Location.from_gps(
*traffic_light_dict['position'])
traffic_lights_list.append(
TrafficLight(
1.0, # confidence
traffic_light_state,
traffic_light_id,
pylot.utils.Transform(
location, pylot.utils.Rotation()) # No rotation given
))
return traffic_lights_list
def get_actors(world):
actor_list = world.get_actors()
tl_actors = actor_list.filter('traffic.traffic_light*')
traffic_lights = [
TrafficLight.from_carla_actor(tl_actor) for tl_actor in tl_actors
]
traffic_stop_actors = actor_list.filter('traffic.stop')
traffic_stops = [
StopSign.from_carla_actor(ts_actor) for ts_actor in traffic_stop_actors
]
speed_limit_actors = actor_list.filter('traffic.speed_limit*')
speed_signs = [
SpeedLimitSign.from_carla_actor(ts_actor)
for ts_actor in speed_limit_actors
]
return (tl_actors, traffic_lights, traffic_stops, speed_signs)
def __convert_to_detected_tl(self, boxes, scores, labels, height, width):
traffic_lights = []
for index in range(len(scores)):
if scores[
index] > self._flags.traffic_light_det_min_score_threshold:
bbox = BoundingBox2D(
int(boxes[index][1] * width), # x_min
int(boxes[index][3] * width), # x_max
int(boxes[index][0] * height), # y_min
int(boxes[index][2] * height) # y_max
)
traffic_lights.append(
TrafficLight(scores[index],
labels[index],
bounding_box=bbox))
return traffic_lights
def extract_data_in_pylot_format(actor_list):
"""Extracts actor information in pylot format from an actor list.
Args:
actor_list (carla.ActorList): An actor list object with all the
simulation actors.
Returns:
A tuple that contains objects for all different types of actors.
"""
# Note: the output will include the ego vehicle as well.
vec_actors = actor_list.filter('vehicle.*')
vehicles = [
Obstacle.from_simulator_actor(vec_actor) for vec_actor in vec_actors
]
person_actors = actor_list.filter('walker.pedestrian.*')
people = [
Obstacle.from_simulator_actor(ped_actor) for ped_actor in person_actors
]
tl_actors = actor_list.filter('traffic.traffic_light*')
traffic_lights = [
TrafficLight.from_simulator_actor(tl_actor) for tl_actor in tl_actors
]
speed_limit_actors = actor_list.filter('traffic.speed_limit*')
speed_limits = [
SpeedLimitSign.from_simulator_actor(ts_actor)
for ts_actor in speed_limit_actors
]
traffic_stop_actors = actor_list.filter('traffic.stop')
traffic_stops = [
StopSign.from_simulator_actor(ts_actor)
for ts_actor in traffic_stop_actors
]
return (vehicles, people, traffic_lights, speed_limits, traffic_stops)
def log_traffic_lights(world):
world_map = world.get_map()
(traffic_lights, _, _, _) = get_actors(world)
tl_colors = [
carla.TrafficLightState.Yellow, carla.TrafficLightState.Green,
carla.TrafficLightState.Red
]
transforms_of_interest = []
for light in traffic_lights:
print("Working for traffic light {}".format(light.id))
# For every traffic light, get the neighbouring lights except the one
# directly opposite.
for offset in range(10, 40, 5):
# Traffic lights have different coordinate systems, hence
# we need to offset y, instead of x and add that to the trigger
# volume location.
offset_loc = pylot.utils.Location(
x=light.trigger_volume.location.x,
y=light.trigger_volume.location.y + offset,
z=light.trigger_volume.location.z)
offset_trans = pylot.utils.Transform(offset_loc,
pylot.utils.Rotation())
# Transform the offset relative to the traffic light.
transform = pylot.utils.Transform.from_carla_transform(
light.get_transform()) * offset_trans
location = transform.location.as_carla_location()
# Get the waypoint nearest to the transform.
w = world_map.get_waypoint(location,
project_to_road=True,
lane_type=carla.LaneType.Driving)
w_rotation = w.transform.rotation
camera_transform = pylot.utils.Transform.from_carla_transform(
w.transform)
camera_transform.location.z += 2.0
transform = camera_transform.as_carla_transform()
transforms_of_interest.append(transform)
# Get the right lanes.
wp_right = w.get_right_lane()
while wp_right and wp_right.lane_type == carla.LaneType.Driving \
and w_rotation == wp_right.transform.rotation:
camera_transform = pylot.utils.Transform.from_carla_transform(
wp_right.transform)
camera_transform.location.z += 2.0
transform = camera_transform.as_carla_transform()
transforms_of_interest.append(transform)
wp_right = wp_right.get_right_lane()
# Get the left lanes.
wp_left = w.get_left_lane()
while wp_left and wp_left.lane_type == carla.LaneType.Driving and \
w_rotation == wp_left.transform.rotation:
camera_transform = pylot.utils.Transform.from_carla_transform(
wp_left.transform)
camera_transform.location.z += 2.0
transform = camera_transform.as_carla_transform()
transforms_of_interest.append(transform)
wp_left = wp_left.get_left_lane()
print("The total number of transforms were: {}".format(
len(transforms_of_interest)))
traffic_lights = [
TrafficLight.from_carla_actor(light) for light in traffic_lights
]
for weather in find_weather_presets():
change_weather(world, weather)
time.sleep(1)
for tl_color in tl_colors:
change_traffic_light_colors(world, tl_color)
world.tick()
time.sleep(1)
log_obstacles(world, transforms_of_interest, traffic_lights,
tl_color, None, None, weather, world_map.name)