def test_extract_box(self):
train_df = parse_train_csv()
box, sample_data_token = extract_single_box(train_df, 0)
sample_token = "92bff46db1dbfc9679edc8091770c4256ac3c027e9f0a9c31dfc4fff41f6f677"
box_from_annotation_token = level5data.get_box(sample_token)
def debug_bounding_box_in_cam_coord():
train_df = parse_train_csv()
data_idx = 0
sample_token, bounding_box = get_train_data_sample_token_and_box(
data_idx, train_df)
object_of_interest_name = ['car', 'pedestrian', 'cyclist']
sample_record = level5data.get('sample', sample_token)
lidar_data_token = sample_record['data']['LIDAR_TOP']
w, l, h = bounding_box.wlh
lwh = np.array([l, w, h])
dummy_bounding_box = bounding_box.copy()
mask, point_clouds_in_box, _, _, image = get_pc_in_image_fov(
lidar_data_token, 'CAM_FRONT', bounding_box)
assert dummy_bounding_box == bounding_box
camera_token = extract_other_sensor_token('CAM_FRONT', lidar_data_token)
bounding_box_sensor_coord = transform_box_from_world_to_sensor_coordinates(
bounding_box,
camera_token,
)
draw_lidar_simple(point_clouds_in_box)
input()
def test_convert_back_world_coordinate(self):
train_df = parse_train_csv()
sample_token, bounding_box = get_train_data_sample_token_and_box(
0, train_df)
# sample_token="db8b47bd4ebdf3b3fb21598bb41bd8853d12f8d2ef25ce76edd4af4d04e49341"
train_sample = level5data.get('sample', sample_token)
cam_token = train_sample['data']['CAM_FRONT']
bounding_box_in_sensor_coord = transform_box_from_world_to_sensor_coordinates(
bounding_box,
cam_token,
)
transform_mtx = get_sensor_to_world_transform_matrix(
sample_token, 'CAM_FRONT')
box_corner_2d, box_corner_3d = transform_bounding_box_to_sensor_coord_and_get_corners(
bounding_box, cam_token, frustum_pointnet_convention=True)
box_center = (box_corner_3d[:, 0] + box_corner_3d[:, 6]) / 2
print(box_center)
print(bounding_box_in_sensor_coord.center)
box_center_h = np.concatenate((box_center, np.ones(1)))
box_center_world = np.dot(transform_mtx, box_center_h.T)
print(box_center_world)
print(bounding_box.center)
def test_get_bounding_box_corners(self):
train_df = parse_train_csv()
sample_token, bounding_box = get_train_data_sample_token_and_box(
0, train_df)
first_train_sample = level5data.get('sample', sample_token)
cam_token = first_train_sample['data']['CAM_FRONT']
box_corners = transform_bounding_box_to_sensor_coord_and_get_corners(
bounding_box, cam_token)
print(box_corners)
# check image
cam_image_file = level5data.get_sample_data_path(cam_token)
cam_image_mtx = imread(cam_image_file)
xmin, xmax, ymin, ymax = get_2d_corners_from_projected_box_coordinates(
box_corners)
fig, ax = plt.subplots(1, 1)
ax.imshow(cam_image_mtx)
ax.plot(box_corners[0, :], box_corners[1, :])
ax.plot([xmin, xmin, xmax, xmax], [ymin, ymax, ymin, ymax])
plt.show()
def plot_demo_lyft_lidar():
train_df = parse_train_csv()
box, sample_data_token = extract_single_box(train_df, 0)
ldp_file_path = level5data.get_sample_data_path(sample_data_token)
lcdp = LidarPointCloud.from_file(ldp_file_path)
draw_lidar_simple(lcdp.points[0:3, :])
input()
def test_get_sample_image_and_transform_box(self):
train_df = parse_train_csv()
sample_token, bounding_box = get_train_data_sample_token_and_box(
0, train_df)
first_train_sample = level5data.get('sample', sample_token)
lidar_data_token = first_train_sample['data']['LIDAR_TOP']
mask, lpc_array_in_cam_coord, filtered_pc_2d, _, image = get_pc_in_image_fov(
lidar_data_token, 'CAM_FRONT', bounding_box)
plt.scatter(lpc_array_in_cam_coord[0, :], lpc_array_in_cam_coord[2, :])
fig, ax = plt.subplots(1, 1)
ax.imshow(image)
print(filtered_pc_2d.shape)
ax.plot(filtered_pc_2d[0, :], filtered_pc_2d[1, :], '.')
plt.show()
def test_transform_image_to_camera_coord(self):
train_df = parse_train_csv()
sample_token, bounding_box = get_train_data_sample_token_and_box(
0, train_df)
first_train_sample = level5data.get('sample', sample_token)
cam_token = first_train_sample['data']['CAM_FRONT']
sd_record = level5data.get("sample_data", cam_token)
cs_record = level5data.get("calibrated_sensor",
sd_record["calibrated_sensor_token"])
cam_intrinsic_mtx = np.array(cs_record["camera_intrinsic"])
box_corners = transform_bounding_box_to_sensor_coord_and_get_corners(
bounding_box, cam_token)
# check)image
cam_image_file = level5data.get_sample_data_path(cam_token)
cam_image_mtx = imread(cam_image_file)
xmin, xmax, ymin, ymax = get_2d_corners_from_projected_box_coordinates(
box_corners)
random_depth = 20
image_center = np.array([[(xmax + xmin) / 2, (ymax + ymin) / 2,
random_depth]]).T
image_center_in_cam_coord = transform_image_to_cam_coordinate(
image_center, cam_token)
self.assertTrue(np.isclose(random_depth,
image_center_in_cam_coord[2:]))
transformed_back_image_center = view_points(image_center_in_cam_coord,
cam_intrinsic_mtx,
normalize=True)
self.assertTrue(
np.allclose(image_center[0:2, :],
transformed_back_image_center[0:2, :]))
import numpy as np
from prepare_lyft_data import level5data, parse_train_csv, \
get_train_data_sample_token_and_box, transform_box_from_world_to_sensor_coordinates
train_df = parse_train_csv()
train_idx = 0
sample_token, train_sample_box = get_train_data_sample_token_and_box(train_idx, train_df)
first_train_sample = level5data.get('sample', sample_token)
sample_data_token = first_train_sample['data']['LIDAR_TOP']
box_in_velo_coordinate = transform_box_from_world_to_sensor_coordinates(train_sample_box, sample_data_token, )
print(box_in_velo_coordinate.center)
print(box_in_velo_coordinate.wlh)