def _get_all_models_info(self):
""" This function is used to gather all the agents, obstacles, bots
spawn on the track. So that the exact location is gathered and icons are drawn
overlaying on top of it. Thus making it prominent.
Returns:
(list, list): model_names (Names of agent, bots, obstacles)
model_imgs (The appropriate image icons for these)
"""
model_names, model_imgs = list(), list()
# Adding all agents to the list
for i, racecar_info in enumerate(self.racecars_info):
model_names.append(racecar_info['name'])
model_imgs.append(utils.get_image(TrackAssetsIconographicPngs.AGENTS_PNG.value[i],
IconographicImageSize.AGENTS_IMAGE_SIZE.value))
# Adding obstacles to the list
num_obstacles = int(rospy.get_param("NUMBER_OF_OBSTACLES", 0))
if num_obstacles:
for i in range(num_obstacles):
model_names.append("obstacle_{}".format(i))
model_imgs.append(utils.get_image(TrackAssetsIconographicPngs.OBSTACLES_PNG.value,
IconographicImageSize.OBSTACLE_IMAGE_SIZE.value))
# Adding bot cars to the list
num_bots = int(rospy.get_param("NUMBER_OF_BOT_CARS", 0))
if num_bots:
for i in range(num_bots):
model_names.append("bot_car_{}".format(i))
model_imgs.append(utils.get_image(TrackAssetsIconographicPngs.BOTS_PNG.value,
IconographicImageSize.BOT_CAR_IMAGE_SIZE.value))
return model_names, model_imgs
def _plot_track_on_gradient(self, gradient_img_path):
""" For the given gradient apply the track iconographic image and use this to apply gradient
on each camera frame. Previously this was done on the top camera which changed every frame. But
with the track iconographic image set static, adding the track on gradient is more optimized.
Arguments:
gradient_img_path (str): Gradient image path
Returns:
(Image): Edited gradient image with track image
"""
gradient_img = utils.get_image(
gradient_img_path, IconographicImageSize.FULL_IMAGE_SIZE.value)
gradient_img = cv2.cvtColor(gradient_img, cv2.COLOR_RGBA2BGRA)
track_icongraphy_scaled = utils.resize_image(self.track_icongraphy_img,
SCALE_RATIO)
track_icongraphy_alpha = track_icongraphy_scaled[:, :, 3] / 255.0
# Track image is placed at the bottom right with some offset (only in leaderboard tracks)
x_min = -(self.track_loc_offset[1] + track_icongraphy_scaled.shape[0])
x_max = gradient_img.shape[0] - self.track_loc_offset[1]
y_min = -(self.track_loc_offset[0] + track_icongraphy_scaled.shape[1])
y_max = gradient_img.shape[1] - self.track_loc_offset[0]
# This is used as the offset for plotting the agent dots
self.track_start_loc = (gradient_img.shape[1] + y_min,
gradient_img.shape[0] + x_min)
for channel in range(0, 4):
gradient_img[x_min:x_max, y_min:y_max, channel] =\
(track_icongraphy_alpha * track_icongraphy_scaled[:, :, channel]) + \
(1 - track_icongraphy_alpha) * (gradient_img[x_min:x_max, y_min:y_max, channel])
return gradient_img
def __init__(self, racecar_name, racecars_info, race_type):
""" This class is used for head to head racing where there are more than one agent
Args:
racecar_name (str): The agent name with 45degree camera view
racecars_info (dict): All the agents information
race_type (str): The type of race. This is used to know if its race type or evaluation
"""
self.racecar_name = racecar_name
self.racecars_info = racecars_info
self.race_type = race_type
# init cv bridge
self.bridge = CvBridge()
# Store the font which we will use to write the phase with
self.amazon_ember_regular_20px = utils.get_font(
'AmazonEmber-Regular', 20)
self.amazon_ember_regular_16px = utils.get_font(
'AmazonEmber-Regular', 16)
self.amazon_ember_heavy_30px = utils.get_font('AmazonEmber-Heavy', 30)
self.amazon_ember_light_18px = utils.get_font('AmazonEmber-Light', 18)
self.amazon_ember_light_20px = utils.get_font('AmazonEmber-Light', 20)
self.amazon_ember_light_italic_20px = utils.get_font(
'AmazonEmber-LightItalic', 20)
self.is_racing = rospy.get_param("VIDEO_JOB_TYPE", "") == "RACING"
self.is_league_leaderboard = rospy.get_param("LEADERBOARD_TYPE",
"") == "LEAGUE"
self.leaderboard_name = rospy.get_param("LEADERBOARD_NAME", "")
# The track image as iconography
self.track_icongraphy_img = utils.get_track_iconography_image()
gradient_img_path = TrackAssetsIconographicPngs.HEAD_TO_HEAD_OVERLAY_PNG_LEAGUE_LEADERBOARD.value \
if self.is_league_leaderboard else TrackAssetsIconographicPngs.HEAD_TO_HEAD_OVERLAY_PNG.value
self.gradient_img = utils.get_image(
gradient_img_path, IconographicImageSize.FULL_IMAGE_SIZE.value)
self.gradient_alpha = self.gradient_img[:, :, 3] / 255.0
self.mp4_video_metrics_srv_list = list()
for racecar_info in self.racecars_info:
agent_name = 'agent' if len(
racecars_info) == 1 else "agent_{}".format(
racecar_info['name'].split("_")[1])
rospy.wait_for_service("/{}/{}".format(agent_name,
"mp4_video_metrics"))
self.mp4_video_metrics_srv_list.append(
ServiceProxyWrapper(
"/{}/{}".format(agent_name, "mp4_video_metrics"),
VideoMetricsSrv))
# Top camera information
top_camera_info = utils.get_top_camera_info()
self.edited_topview_pub = rospy.Publisher('/deepracer/topview_stream',
ROSImg,
queue_size=1)
self.top_view_graphics = TopViewGraphics(
top_camera_info.horizontal_fov, top_camera_info.padding_pct,
top_camera_info.image_width, top_camera_info.image_height,
racecars_info)
def __init__(self, total_sectors):
"""initialize Finish state with a finish time for how long to display check flag
Args:
total_sectors (int): total number of sectors
"""
LOG.info("[virtual event]: video edit state at {}".format(self))
self._icon_image = utils.get_image(
VirtualEventIconographicPngs.FINISH.value)
self._icon_image = cv2.cvtColor(self._icon_image, cv2.COLOR_RGBA2BGRA)
self._sectors = [
SECTOR_X_FORMAT.format(idx + 1) for idx in range(total_sectors)
]
def __init__(self, racecar_info, race_type):
""" Initializing the required data for the head to bot, time-trail. This is used for single agent
Arguments:
racecars_info (list): list of dict having information of the agent
race_type (str): Since this class is reused for all the different race_type
"""
self.racecar_info = racecar_info
self.race_type = race_type
# Store the font which we will use to write the phase with
self.amazon_ember_regular_20px = utils.get_font(
'AmazonEmber-Regular', 20)
self.amazon_ember_regular_16px = utils.get_font(
'AmazonEmber-Regular', 16)
self.amazon_ember_heavy_30px = utils.get_font('AmazonEmber-Heavy', 30)
self.amazon_ember_light_18px = utils.get_font('AmazonEmber-Light', 18)
self.amazon_ember_light_20px = utils.get_font('AmazonEmber-Light', 20)
self.amazon_ember_light_italic_20px = utils.get_font(
'AmazonEmber-LightItalic', 20)
self.is_racing = rospy.get_param("VIDEO_JOB_TYPE", "") == "RACING"
self.is_league_leaderboard = rospy.get_param("LEADERBOARD_TYPE",
"") == "LEAGUE"
self.leaderboard_name = rospy.get_param("LEADERBOARD_NAME", "")
# The track image as iconography
self.track_icongraphy_img = utils.get_track_iconography_image()
gradient_img_path = TrackAssetsIconographicPngs.OBSTACLE_OVERLAY_PNG_LEAGUE_LEADERBOARD.value \
if self.is_league_leaderboard else TrackAssetsIconographicPngs.OBSTACLE_OVERLAY_PNG.value
self.gradient_img = utils.get_image(
gradient_img_path, IconographicImageSize.FULL_IMAGE_SIZE.value)
self.gradient_alpha = self.gradient_img[:, :, 3] / 255.0
# Subscribing to the agent metrics
rospy.wait_for_service("/agent/mp4_video_metrics")
self.mp4_video_metrics_srv = ServiceProxyWrapper(
"/agent/mp4_video_metrics", VideoMetricsSrv)
# Top camera information
top_camera_info = utils.get_top_camera_info()
self.edited_topview_pub = rospy.Publisher('/deepracer/topview_stream',
ROSImg,
queue_size=1)
self.top_view_graphics = TopViewGraphics(
top_camera_info.horizontal_fov, top_camera_info.padding_pct,
top_camera_info.image_width, top_camera_info.image_height,
racecar_info)
def _edit_major_cv_image(self, major_cv_image):
""" Apply all the editing for the Major 45degree camera image
Args:
major_cv_image (Image): Image straight from the camera
Returns:
Image: Edited main camera image
"""
# Applying gradient to whole major image and then writing text
major_cv_image = utils.apply_gradient(major_cv_image,
self.gradient_img,
self.gradient_alpha)
# Subscribing to the agent metrics
mp4_video_metrics_info = list()
for racecar_info, mp4_video_metrics_srv in zip(
self.racecars_info, self.mp4_video_metrics_srv_list):
mp4_video_metrics = mp4_video_metrics_srv(VideoMetricsSrvRequest())
mp4_video_metrics_info.append(mp4_video_metrics)
# Adding display name to the image
agents_speed = 0
agent_done = False
for i, racecar_info in enumerate(self.racecars_info):
loc_x, loc_y = XYPixelLoc.MULTI_AGENT_DISPLAY_NAME_LOC.value[i][
0], XYPixelLoc.MULTI_AGENT_DISPLAY_NAME_LOC.value[i][1]
# Display name (Racer name/Model name)
display_name = racecar_info['display_name']
display_name_txt = display_name if len(
display_name) < 15 else "{}...".format(display_name[:15])
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=display_name_txt,
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Lap Counter
loc_y += 30
total_laps = rospy.get_param("NUMBER_OF_TRIALS", 0)
current_lap = int(mp4_video_metrics_info[i].lap_counter) + 1
lap_counter_text = "{}/{}".format(current_lap, total_laps)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=lap_counter_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_heavy_30px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Reset counter
loc_y += 45
reset_counter_text = "Reset | {}".format(
mp4_video_metrics_info[i].reset_counter)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=reset_counter_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_18px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
if self.racecar_name == racecar_info['name']:
agents_speed = mp4_video_metrics_info[i].throttle
# The race is complete when total lap is same as current lap and done flag is set
agent_done = agent_done or (mp4_video_metrics_info[i].done and
(current_lap == int(total_laps)))
# Speed
loc_x, loc_y = XYPixelLoc.SPEED_EVAL_LOC.value
if self.is_league_leaderboard:
loc_x, loc_y = XYPixelLoc.SPEED_LEADERBOARD_LOC.value
speed_text = "{} m/s".format(
utils.get_speed_formatted_str(agents_speed))
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=speed_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Leaderboard name
if self.is_league_leaderboard:
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=self.leaderboard_name,
loc=XYPixelLoc.LEADERBOARD_NAME_LOC.value,
font=self.amazon_ember_regular_16px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Evaluation type
loc_x, loc_y = XYPixelLoc.RACE_TYPE_EVAL_LOC.value
if self.is_league_leaderboard:
loc_x, loc_y = XYPixelLoc.RACE_TYPE_RACE_LOC.value
race_text = "race" if self.is_racing else "evaluation"
evaluation_type_txt = "{} {}".format(
RACE_TYPE_TO_VIDEO_TEXT_MAPPING[self.race_type], race_text)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=evaluation_type_txt,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_italic_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# total_evaluation_time (Race time)
loc_x, loc_y = XYPixelLoc.MULTI_AGENT_EVAL_TIME.value
total_eval_milli_seconds = mp4_video_metrics_info[
0].total_evaluation_time
time_delta = datetime.timedelta(milliseconds=total_eval_milli_seconds)
total_eval_time_text = "Race | {}".format(
utils.milliseconds_to_timeformat(time_delta))
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=total_eval_time_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_18px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# AWS Deepracer logo at the bottom for the community leaderboard
if self.is_league_leaderboard:
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=AWS_DEEPRACER_WATER_MARK,
loc=XYPixelLoc.AWS_DEEPRACER_WATER_MARK_LOC.value,
font=self.amazon_ember_regular_16px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Check if the done flag is set and set the banner appropriately
if agent_done:
# When the cv2 text is written, it automatically drops the alpha value of the image
rel_y_offset = XYPixelLoc.TRACK_IMG_WITH_OFFSET_LOC.value[
1] if self.is_league_leaderboard else 0
major_cv_image = cv2.cvtColor(major_cv_image, cv2.COLOR_RGB2RGBA)
racecomplete_image = utils.get_image(
TrackAssetsIconographicPngs.RACE_COMPLETE_OVERLAY_PNG.value,
IconographicImageSize.RACE_COMPLETE_IMAGE_SIZE.value)
x_offset = major_cv_image.shape[
1] - racecomplete_image.shape[1] // 2
y_offset = major_cv_image.shape[
0] - RACE_COMPLETE_Y_OFFSET - rel_y_offset - racecomplete_image.shape[
0] // 2
major_cv_image = utils.plot_rectangular_image_on_main_image(
major_cv_image, racecomplete_image, (x_offset, y_offset))
return major_cv_image
def __init__(self, current_sector=0):
"""initialize Run state and log
Args:
current_sector (int): current sector index 0-based.
"""
self._total_sectors = int(rospy.get_param("NUM_SECTORS", "3"))
if self._total_sectors == 0:
log_and_exit(
"[virtual event]: Virtual event run state with 0 total sectors. \
This needs to be at least 1",
SIMAPP_SIMULATION_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_500)
# current sector index 0 is sector1 and so on so forth
self._current_sector = current_sector % self._total_sectors
self._target_progress = (100.00 / self._total_sectors) * (
self._current_sector + 1)
self._total_laps = int(rospy.get_param("NUMBER_OF_TRIALS", 3))
self._race_duration = int(
rospy.get_param("RACE_DURATION", DEFAULT_RACE_DURATION)) * 1000
# VirtualEventBestSectorTime S3 upload instance
# use the s3 bucket and prefix for yaml file stored as environment variable because
# here is SimApp use only. For virtual event there is no s3 bucket and prefix past
# through yaml file. All are past through sqs. For simplicity, reuse the yaml s3 bucket
# and prefix environment variable.
self._virtual_event_best_sector_time = VirtualEventBestSectorTime(
bucket=os.environ.get("YAML_S3_BUCKET", ''),
s3_key=get_s3_key(os.environ.get("YAML_S3_PREFIX", ''),
SECTOR_TIME_S3_POSTFIX),
region_name=os.environ.get("APP_REGION", "us-east-1"),
local_path=SECTOR_TIME_LOCAL_PATH)
# Go icon image
self._icon_image = utils.get_image(
VirtualEventIconographicPngs.GO.value)
self._icon_image = cv2.cvtColor(self._icon_image, cv2.COLOR_RGBA2BGRA)
# init number of sectors to plot
# for 3 sectors example
# if racer is at sector 1 (idx 0): plot sector 1, 2, and 3
# if racer is at sector 2 (idx 1): plot sector 1
# if racer is at sector 3 (idx 2): plot sector 1, and 2
num_sectors_to_plot = self._current_sector
if self._current_sector == 0:
num_sectors_to_plot = self._total_sectors
# init number of sectors to plot
self._sectors = [
SECTOR_X_FORMAT.format(idx + 1)
for idx in range(num_sectors_to_plot)
]
# sector format string for best session, best personal, and current personal
self._best_session_format = SECTOR_TIME_FORMAT_DICT[
TrackSectorTime.BEST_SESSION]
self._best_personal_format = SECTOR_TIME_FORMAT_DICT[
TrackSectorTime.BEST_PERSONAL]
self._current_personal_format = SECTOR_TIME_FORMAT_DICT[
TrackSectorTime.CURRENT_PERSONAL]
LOG.info(
"[virtual event]: video edit state at {} for sector {} with target progress {}\
".format(self, self._current_sector + 1, self._target_progress))
def __init__(self, racecar_name, racecars_info, race_type):
""" This class is used for head to head racing where there are more than one agent
Args:
racecar_name (str): The agent name with 45degree camera view
racecars_info (dict): All the agents information
race_type (str): The type of race. This is used to know if its race type or evaluation
"""
self.racecar_name = racecar_name
self.racecars_info = racecars_info
racecar_index = get_racecar_idx(racecar_name)
self.racecar_index = racecar_index if racecar_index else 0
self.race_type = race_type
# Store the font which we will use to write the phase with
self.formula1_display_regular_12px = utils.get_font(
'Formula1-Display-Regular', 12)
self.formula1_display_regular_14px = utils.get_font(
'Formula1-Display-Regular', 14)
self.formula1_display_regular_16px = utils.get_font(
'Formula1-Display-Regular', 16)
self.formula1_display_wide_12px = utils.get_font(
'Formula1-Display-Wide', 12)
self.formula1_display_bold_16px = utils.get_font(
'Formula1-Display-Bold', 16)
self.total_laps = int(rospy.get_param("NUMBER_OF_TRIALS", 0))
self.is_league_leaderboard = rospy.get_param("LEADERBOARD_TYPE",
"") == "LEAGUE"
self.leaderboard_name = rospy.get_param("LEADERBOARD_NAME", "")
# The track image as iconography
self.track_icongraphy_img = utils.get_track_iconography_image()
# Track image offset
self.track_loc_offset = XYPixelLoc.TRACK_IMG_WITHOUT_OFFSET_LOC.value
# Default image of top view
gradient_default_img_path = TrackAssetsIconographicPngs.F1_OVERLAY_DEFAULT_PNG.value
self.gradient_default_img = self._plot_track_on_gradient(
gradient_default_img_path)
self.gradient_default_alpha_rgb_mul, self.one_minus_gradient_default_alpha = utils.get_gradient_values(
self.gradient_default_img)
# Midway track gradient
gradient_midway_img_path = TrackAssetsIconographicPngs.F1_OVERLAY_MIDWAY_PNG.value
self.gradient_midway_img = self._plot_track_on_gradient(
gradient_midway_img_path)
self.gradient_midway_alpha_rgb_mul, self.one_minus_gradient_midway_alpha = utils.get_gradient_values(
self.gradient_midway_img)
# Finisher track gradient
gradient_finisher_img_path = TrackAssetsIconographicPngs.F1_OVERLAY_FINISHERS_PNG.value
self.gradient_finisher_img = self._plot_track_on_gradient(
gradient_finisher_img_path)
self.gradient_finisher_alpha_rgb_mul, self.one_minus_gradient_finisher_alpha = utils.get_gradient_values(
self.gradient_finisher_img)
# Top camera gradient
num_racers = len(self.racecars_info)
if num_racers <= 8:
# TODO: Add one box image and use 1 box image if number of racers are <= 4.
gradient_top_camera_img_path = TrackAssetsIconographicPngs.F1_OVERLAY_TOPVIEW_2BOX_PNG.value
elif num_racers <= 12:
gradient_top_camera_img_path = TrackAssetsIconographicPngs.F1_OVERLAY_TOPVIEW_3BOX_PNG.value
else:
raise Exception(
"More than 12 racers are not supported for Grand Prix")
gradient_top_camera_img = utils.get_image(
gradient_top_camera_img_path,
IconographicImageSize.FULL_IMAGE_SIZE.value)
gradient_top_camera_img = cv2.cvtColor(gradient_top_camera_img,
cv2.COLOR_RGBA2BGRA)
self.gradient_top_camera_alpha_rgb_mul, self.one_minus_gradient_top_camera_alpha = utils.get_gradient_values(
gradient_top_camera_img)
# Top camera information
top_camera_info = utils.get_top_camera_info()
self.edited_topview_pub = rospy.Publisher('/deepracer/topview_stream',
ROSImg,
queue_size=1)
self.top_view_graphics = TopViewGraphics(
top_camera_info.horizontal_fov, top_camera_info.padding_pct,
top_camera_info.image_width, top_camera_info.image_height,
racecars_info, race_type)
self.hex_car_colors = [
val['racecar_color'].split('_')[-1] for val in racecars_info
]
self._racer_color_code_rect_img = list()
self._racer_color_code_slash_img = list()
for car_color in self.hex_car_colors:
# Rectangular png of racers
racer_color_code_rect = "{}_{}".format(
TrackAssetsIconographicPngs.F1_AGENTS_RECT_DISPLAY_ICON_PNG.
value, car_color)
self._racer_color_code_rect_img.append(
utils.get_image(
racer_color_code_rect, IconographicImageSize.
F1_RACER_RECT_DISPLAY_ICON_SIZE.value))
# Slash png of racers
racer_color_code_slash = "{}_{}".format(
TrackAssetsIconographicPngs.F1_AGENTS_SLASH_DISPLAY_ICON_PNG.
value, car_color)
racer_color_code_slash_img = utils.get_image(
racer_color_code_slash,
IconographicImageSize.F1_RACER_SLASH_DISPLAY_ICON_SIZE.value)
self._racer_color_code_slash_img.append(
cv2.cvtColor(racer_color_code_slash_img, cv2.COLOR_RGBA2BGRA))
def _get_all_models_info(self):
""" This function is used to gather all the agents, obstacles, bots
spawn on the track. So that the exact location is gathered and icons are drawn
overlaying on top of it. Thus making it prominent.
Returns:
(dict): "object_imgs" - Appropriate image icons for boxes and bots
"agent_imgs" - Appropriate image icons for agents
"agent_num_imgs" - Appropriate image icons for agents number
"""
object_imgs = list()
agent_imgs = list()
agent_num_imgs = list()
# Adding obstacles to the list
num_obstacles = int(rospy.get_param("NUMBER_OF_OBSTACLES", 0))
if num_obstacles:
# Other agents also come as object_locations so first plot all the obstacles and
# Then overlay agents on top.
for i in range(num_obstacles):
object_imgs.append(
utils.get_image(
TrackAssetsIconographicPngs.OBSTACLES_PNG.value,
IconographicImageSize.OBSTACLE_IMAGE_SIZE.value))
# Adding bot cars to the list
num_bots = int(rospy.get_param("NUMBER_OF_BOT_CARS", 0))
if num_bots:
# Other agents also come as object_locations so first plot all the obstacles and
# Then overlay agents on top.
for i in range(num_bots):
object_imgs.append(
utils.get_image(
TrackAssetsIconographicPngs.BOTS_PNG.value,
IconographicImageSize.BOT_CAR_IMAGE_SIZE.value))
# Adding all agents to the list
agents_png = TrackAssetsIconographicPngs.VIRTUAL_EVENT_AGENTS_PNG.value \
if self._is_virtual_event else TrackAssetsIconographicPngs.AGENTS_PNG.value
agent_size = IconographicImageSize.VIRTUAL_EVENT_AGENTS_IMAGE_SIZE.value \
if self._is_virtual_event else IconographicImageSize.AGENTS_IMAGE_SIZE.value
for i, _ in enumerate(self.racecars_info):
# If the number of racecars are greater than 2
if self.is_f1_race_type:
hex_car_color = self.racecars_info[i]['racecar_color'].split(
'_')[-1]
racer_number = int(self.racecars_info[i]['name'].split("_")[1])
agents_round_img = "{}_{}".format(
TrackAssetsIconographicPngs.F1_AGENTS_PNG.value,
hex_car_color)
agent_imgs.append(
utils.get_image(
agents_round_img,
IconographicImageSize.F1_AGENTS_IMAGE_SIZE.value))
agent_number_img_name = "{}_{}".format(
TrackAssetsIconographicPngs.F1_AGENTS_NUM_PNG.value,
racer_number + 1)
agent_num_imgs.append(
utils.get_image(
agent_number_img_name,
IconographicImageSize.F1_AGENTS_IMAGE_SIZE.value))
else:
agent_imgs.append(
utils.get_image(
TrackAssetsIconographicPngs.AGENTS_PNG.value[i % 2],
IconographicImageSize.AGENTS_IMAGE_SIZE.value))
return {
ModelImgNames.OBJECT_IMGS.value: object_imgs,
ModelImgNames.AGENT_IMGS.value: agent_imgs,
ModelImgNames.AGENT_NUM_IMGS.value: agent_num_imgs
}
def _edit_major_cv_image(self, major_cv_image, mp4_video_metrics_info):
""" Apply all the editing for the Major 45degree camera image
Args:
major_cv_image (Image): Image straight from the camera
Returns:
Image: Edited main camera image
"""
# Applying gradient to whole major image and then writing text
major_cv_image = utils.apply_gradient(major_cv_image,
self.gradient_alpha_rgb_mul,
self.one_minus_gradient_alpha)
# Top left location of the picture
loc_x, loc_y = XYPixelLoc.SINGLE_AGENT_DISPLAY_NAME_LOC.value
# Display name (Racer name/Model name)
display_name = self.racecar_info[self.racecar_index]['display_name']
display_name_txt = display_name if len(
display_name) < 15 else "{}...".format(display_name[:15])
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=display_name_txt,
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Lap Counter
loc_y += 30
total_laps = rospy.get_param("NUMBER_OF_TRIALS", 0)
current_lap = min(
int(mp4_video_metrics_info[self.racecar_index].lap_counter) + 1,
total_laps)
lap_counter_text = "{}/{}".format(current_lap, total_laps)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=lap_counter_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_heavy_30px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# total_evaluation_time (Race time)
loc_y += 45
total_eval_milli_seconds = mp4_video_metrics_info[
self.racecar_index].total_evaluation_time
time_delta = datetime.timedelta(milliseconds=total_eval_milli_seconds)
total_eval_time_text = "Race | {}".format(
utils.milliseconds_to_timeformat(time_delta))
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=total_eval_time_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_18px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Reset counter
loc_y += 25
reset_counter_text = "Reset | {}".format(
mp4_video_metrics_info[self.racecar_index].reset_counter)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=reset_counter_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_18px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Speed
loc_x, loc_y = XYPixelLoc.SPEED_EVAL_LOC.value
if self.is_league_leaderboard:
loc_x, loc_y = XYPixelLoc.SPEED_LEADERBOARD_LOC.value
speed_text = "{} m/s".format(
utils.get_speed_formatted_str(
mp4_video_metrics_info[self.racecar_index].throttle))
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=speed_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Leaderboard name
if self.is_league_leaderboard:
loc_x, loc_y = XYPixelLoc.LEADERBOARD_NAME_LOC.value
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=self.leaderboard_name,
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_16px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Evaluation type
loc_x, loc_y = XYPixelLoc.RACE_TYPE_EVAL_LOC.value
if self.is_league_leaderboard:
loc_x, loc_y = XYPixelLoc.RACE_TYPE_RACE_LOC.value
race_text = "race" if self.is_racing else "evaluation"
evaluation_type_txt = "{} {}".format(
RACE_TYPE_TO_VIDEO_TEXT_MAPPING[self.race_type], race_text)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=evaluation_type_txt,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_italic_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# AWS Deepracer logo at the bottom for the community leaderboard
if self.is_league_leaderboard:
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=AWS_DEEPRACER_WATER_MARK,
loc=XYPixelLoc.AWS_DEEPRACER_WATER_MARK_LOC.value,
font=self.amazon_ember_regular_16px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Check if the done flag is set and set the banner appropriately
if mp4_video_metrics_info[self.racecar_index].done and (
int(total_laps) >= current_lap):
# When the cv2 text is written, it automatically drops the alpha value of the image
rel_y_offset = XYPixelLoc.TRACK_IMG_WITH_OFFSET_LOC.value[
1] if self.is_league_leaderboard else 0
racecomplete_image = utils.get_image(
TrackAssetsIconographicPngs.RACE_COMPLETE_OVERLAY_PNG.value,
IconographicImageSize.RACE_COMPLETE_IMAGE_SIZE.value)
x_offset = major_cv_image.shape[
1] - racecomplete_image.shape[1] // 2
y_offset = major_cv_image.shape[
0] - RACE_COMPLETE_Y_OFFSET - rel_y_offset - racecomplete_image.shape[
0] // 2
major_cv_image = utils.plot_rectangular_image_on_main_image(
major_cv_image, racecomplete_image, (x_offset, y_offset))
major_cv_image = cv2.cvtColor(major_cv_image, cv2.COLOR_RGB2BGRA)
return major_cv_image
def __init__(self, racecar_name, racecar_info, race_type):
""" Initializing the required data for the head to bot, time-trail. This is used for single agent
Arguments:
racecar_name (str): racecar name in string
racecars_info (list): list of dict having information of the agent
race_type (str): Since this class is reused for all the different race_type
"""
# race duration in milliseconds
self._world_name = rospy.get_param("WORLD_NAME")
self.num_sectors = int(rospy.get_param("NUM_SECTORS", "3"))
self.race_duration = int(
rospy.get_param("RACE_DURATION", DEFAULT_RACE_DURATION)) * 1000
self.racecar_info = racecar_info
self.race_type = race_type
racecar_index = get_racecar_idx(racecar_name)
self.racecar_index = racecar_index if racecar_index else 0
# Store the font which we will use to write the phase with
self.amazon_ember_regular_28px = utils.get_font(
'AmazonEmber-Regular', 28)
self.amazon_ember_regular_14px = utils.get_font(
'AmazonEmber-Regular', 14)
# The track image as iconography
self.track_icongraphy_img = utils.get_track_iconography_image()
# Track image offset
self.track_loc_offset = VirtualEventXYPixelLoc.TRACK_IMG_VIRTUAL_EVENT_LOC.value
self._track_x_min = None
self._track_x_max = None
self._track_y_min = None
self._track_y_max = None
# Gradient overlay image with track and virtual event mock
gradient_img_path = VirtualEventIconographicPngs.OVERLAY_PNG.value
self.gradient_img = self._plot_track_on_gradient(gradient_img_path)
# Time remaining text
loc_x, loc_y = VirtualEventXYPixelLoc.TIME_REMAINING_TEXT.value
self.gradient_img = utils.write_text_on_image(
image=self.gradient_img,
text="TIME REMAINING",
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_14px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Speed text
loc_x, loc_y = VirtualEventXYPixelLoc.SPEED_TEXT.value
self.gradient_img = utils.write_text_on_image(
image=self.gradient_img,
text="m/s",
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_14px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Reset text
loc_x, loc_y = VirtualEventXYPixelLoc.RESET_TEXT.value
self.gradient_img = utils.write_text_on_image(
image=self.gradient_img,
text="RESET",
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_14px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# current lap time text
loc_x, loc_y = VirtualEventXYPixelLoc.CURRENT_LAP_TIME_TEXT.value
self.gradient_img = utils.write_text_on_image(
image=self.gradient_img,
text="CURRENT LAP TIME",
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_14px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# best lap time text
loc_x, loc_y = VirtualEventXYPixelLoc.BEST_LAP_TIME_TEXT.value
self.gradient_img = utils.write_text_on_image(
image=self.gradient_img,
text="BEST LAP TIME",
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_14px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# apply graident
self.gradient_alpha_rgb_mul, self.one_minus_gradient_alpha = utils.get_gradient_values(
self.gradient_img)
# Top camera information
top_camera_info = utils.get_top_camera_info()
self.top_view_graphics = TopViewGraphics(
top_camera_info.horizontal_fov,
top_camera_info.padding_pct,
top_camera_info.image_width,
top_camera_info.image_height,
racecar_info,
is_virtual_event=True)
# virtual event image editting state machine
self._image_edit_fsm = FSM(initial_state=VirtualEventWaitState())
# if best sector time download from s3 failed. Then, initialize best sector time as None
# and not display sector color
self._sector_times = {}
# declare sector images
self._sectors_img_dict = {}
for idx in range(self.num_sectors):
sector = SECTOR_X_FORMAT.format(idx + 1)
sector_color_img_dict = utils.init_sector_img_dict(
world_name=self._world_name, sector=sector)
self._sectors_img_dict[sector] = sector_color_img_dict
# use the s3 bucket and prefix for yaml file stored as environment variable because
# here is SimApp use only. For virtual event there is no s3 bucket and prefix past
# through yaml file. All are past through sqs. For simplicity, reuse the yaml s3 bucket
# and prefix environment variable.
self._virtual_event_best_sector_time = VirtualEventBestSectorTime(
bucket=os.environ.get("YAML_S3_BUCKET", ''),
s3_key=get_s3_key(os.environ.get("YAML_S3_PREFIX", ''),
SECTOR_TIME_S3_POSTFIX),
region_name=os.environ.get("APP_REGION", "us-east-1"),
local_path=SECTOR_TIME_LOCAL_PATH)
self._sector_times.update(
self._virtual_event_best_sector_time.get_sector_time(
num_sectors=self.num_sectors))
# declare default best personal and current persoanl time to inf
for idx in range(self.num_sectors):
sector = SECTOR_X_FORMAT.format(idx + 1)
self._sector_times[SECTOR_TIME_FORMAT_DICT[
TrackSectorTime.BEST_PERSONAL].format(sector)] = float("inf")
self._sector_times[SECTOR_TIME_FORMAT_DICT[
TrackSectorTime.CURRENT_PERSONAL].format(sector)] = float(
"inf")
self._curr_lap_time = 0
self._last_eval_time = 0
self._curr_progress = 0
self._last_progress = 0
self._current_lap = 1
# Initializing the fader behaviour to pre-compute the gradient values
final_fading_image = utils.get_image(
VirtualEventIconographicPngs.FINAL_FADING_IMAGE_50ALPHA.value,
IconographicImageSize.FULL_IMAGE_SIZE.value)
final_fading_image = cv2.cvtColor(final_fading_image,
cv2.COLOR_RGBA2BGRA)
self._fader_obj = Fader(
final_fading_image,
fading_min_percent=VirtualEventFader.FADING_MIN_PERCENT.value,
fading_max_percent=VirtualEventFader.FADING_MAX_PERCENT.value,
num_frames=VirtualEventFader.NUM_FRAMES.value)
def _edit_major_cv_image(self, major_cv_image):
""" Apply all the editing for the Major 45degree camera image
Args:
major_cv_image (Image): Image straight from the camera
Returns:
Image: Edited main camera image
"""
# Applying gradient to whole major image and then writing text
major_cv_image = utils.apply_gradient(major_cv_image,
self.gradient_img,
self.gradient_alpha)
# Subscribing to the agent metrics
mp4_video_metrics_info = list()
for racecar_info, mp4_video_metrics_srv in zip(
self.racecars_info, self.mp4_video_metrics_srv_list):
mp4_video_metrics = mp4_video_metrics_srv(VideoMetricsSrvRequest())
mp4_video_metrics_info.append(mp4_video_metrics)
# Adding display name to the image
display_name_loc = [(10, 10), (450, 10)]
agents_speed = 0
agent_done = False
for i, racecar_info in enumerate(self.racecars_info):
loc_x, loc_y = display_name_loc[i][0], display_name_loc[i][1]
# Display name (Racer name/Model name)
display_name = racecar_info['display_name']
display_name_txt = display_name if len(
display_name) < 15 else "{}...".format(display_name[:15])
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=display_name_txt,
loc=(loc_x, loc_y),
font=self.amazon_ember_regular_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Lap Counter
loc_y += 30
total_laps = rospy.get_param("NUMBER_OF_TRIALS", 0)
lap_counter_text = "{}/{}".format(
int(mp4_video_metrics_info[i].lap_counter), total_laps)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=lap_counter_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_heavy_30px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Reset counter
loc_y += 45
reset_counter_text = "Reset | {}".format(
mp4_video_metrics_info[i].reset_counter)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=reset_counter_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_18px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
if self.racecar_name == racecar_info['name']:
agents_speed = mp4_video_metrics_info[i].throttle
agent_done = agent_done or mp4_video_metrics_info[i].done
# Speed
loc_x, loc_y = 10, 420
speed_text = "{} m/s".format(
utils.get_speed_formatted_str(agents_speed))
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=speed_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Evaluation type
loc_y += 25
# TODO - Show text based on whether its a race or customer run evaluation
race_text = "race"
evaluation_type_txt = "{} {}".format(
RACE_TYPE_TO_VIDEO_TEXT_MAPPING[self.race_type], race_text)
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=evaluation_type_txt,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_italic_20px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# total_evaluation_time (Race time)
loc_x, loc_y = 240, 10
total_eval_milli_seconds = mp4_video_metrics_info[
0].total_evaluation_time
time_delta = datetime.timedelta(milliseconds=total_eval_milli_seconds)
total_eval_time_text = "Race | {}".format(
utils.milliseconds_to_timeformat(time_delta))
major_cv_image = utils.write_text_on_image(
image=major_cv_image,
text=total_eval_time_text,
loc=(loc_x, loc_y),
font=self.amazon_ember_light_18px,
font_color=RaceCarColorToRGB.White.value,
font_shadow_color=RaceCarColorToRGB.Black.value)
# Check if the done flag is set and set the banner appropriately
if agent_done:
# When the cv2 text is written, it automatically drops the alpha value of the image
major_cv_image = cv2.cvtColor(major_cv_image, cv2.COLOR_RGB2RGBA)
racecomplete_image = utils.get_image(
TrackAssetsIconographicPngs.RACE_COMPLETE_OVERLAY_PNG.value,
IconographicImageSize.RACE_COMPLETE_IMAGE_SIZE.value)
x_offset = major_cv_image.shape[
1] - racecomplete_image.shape[1] // 2
y_offset = major_cv_image.shape[
0] - 180 - racecomplete_image.shape[0] // 2
major_cv_image = utils.plot_rectangular_image_on_main_image(
major_cv_image, racecomplete_image, (x_offset, y_offset))
return major_cv_image
