# Exercise C2-3-2 : Transform metric point coordinates to BEV space
l2_exercises.pcl_to_bev(cropped_pcl, configs)
# Example C2-3-3 : Minimum and maximum intensity
# l2_examples.min_max_intensity(lidar_pcl)
# Example C2-4-2 : count total no. of vehicles and vehicles that are difficult to track
# l2_examples.count_vehicles(frame)
# Example C2-4-3 : Display label bounding boxes on top of BEV map
# lidar_bev = load_object_from_file(results_fullpath, data_filename, 'lidar_bev', cnt_frame)
# lidar_bev_labels = l2_examples.render_bb_over_bev(lidar_bev, frame.laser_labels, configs)
# Example C2-4-4 : Display detected objects on top of BEV map
detections = load_object_from_file(
results_fullpath, data_filename,
'detections_' + configs.model + '_' + str(configs.conf_thresh),
cnt_frame)
l2_examples.render_obj_over_bev(detections, lidar_bev_labels, configs,
True)
# Exercise C2-4-5 : Compute precision and recall (part 1/2 - remove comments only, no action inside functions required)
det_performance = load_object_from_file(
results_fullpath, data_filename, 'det_performance_' +
configs.model + '_' + str(configs.conf_thresh), cnt_frame)
det_performance_all.append(
det_performance
) # store all evaluation results in a list for performance assessme
#######
####### LESSON 2 EXERCISES & EXAMPLES END #######
lidar_name = dataset_pb2.LaserName.TOP
camera_name = dataset_pb2.CameraName.FRONT
lidar_calibration = waymo_utils.get(frame.context.laser_calibrations,
lidar_name)
camera_calibration = waymo_utils.get(frame.context.camera_calibrations,
camera_name)
if 'load_image' in exec_list:
image = tools.extract_front_camera_image(frame)
## Compute lidar point-cloud from range image
if 'pcl_from_rangeimage' in exec_list:
print('computing point-cloud from lidar range image')
lidar_pcl = tools.pcl_from_range_image(frame, lidar_name)
else:
print('loading lidar point-cloud from result file')
lidar_pcl = load_object_from_file(results_fullpath, data_filename,
'lidar_pcl', cnt_frame)
## Compute lidar birds-eye view (bev)
if 'bev_from_pcl' in exec_list:
print('computing birds-eye view from lidar pointcloud')
lidar_bev = pcl.bev_from_pcl(lidar_pcl, configs_det)
else:
print('loading birds-eve view from result file')
lidar_bev = load_object_from_file(results_fullpath, data_filename,
'lidar_bev', cnt_frame)
## 3D object detection
if (configs_det.use_labels_as_objects == True):
print('using groundtruth labels as objects')
detections = tools.convert_labels_into_objects(
frame.laser_labels, configs_det)
lidar_name = dataset_pb2.LaserName.TOP
camera_name = dataset_pb2.CameraName.FRONT
lidar_calibration = waymo_utils.get(frame.context.laser_calibrations,
lidar_name)
camera_calibration = waymo_utils.get(frame.context.camera_calibrations,
camera_name)
if 'load_image' in exec_list:
image = tools.extract_front_camera_image(frame)
## Compute lidar point-cloud from range image
if 'pcl_from_rangeimage' in exec_list:
print('computing point-cloud from lidar range image')
lidar_pcl = tools.pcl_from_range_image(frame, lidar_name)
else:
print('loading lidar point-cloud from result file')
lidar_pcl = load_object_from_file(results_fullpath, data_filename,
'lidar_pcl', cnt_frame)
## Compute lidar birds-eye view (bev)
if 'bev_from_pcl' in exec_list:
print('computing birds-eye view from lidar pointcloud')
lidar_bev = pcl.bev_from_pcl(lidar_pcl, configs_det)
else:
print('loading birds-eve view from result file')
lidar_bev = load_object_from_file(results_fullpath, data_filename,
'lidar_bev', cnt_frame)
## 3D object detection
if (configs_det.use_labels_as_objects == True):
print('using groundtruth labels as objects')
detections = tools.convert_labels_into_objects(
frame.laser_labels, configs_det)