def try_undistort(desired_count):
undist = cc.CameraConverter()
bagdir = '/data/bags/didi-round2/release/car/training/suburu_leading_at_distance'
bt = mb.find_bag_tracklets(bagdir, '/data/tracklets')
multi = mb.MultiBagStream(bt, ns.generate_numpystream)
generator = multi.generate(infinite = False)
count = 0
output_count = 0
for numpydata in generator:
im = numpydata.image
frame_idx, obs = numpydata.obs
im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
undistorted = undist.undistort_image(im)
if count % 25 == 0:
cv2.imwrite('/data/dev/camera/orig_{}.png'.format(count), im)
# Print center.
img_point = undist.project_point(obs.position)
cv2.circle(undistorted, (int(img_point[0]), int(img_point[1])), radius = 5, color = (255, 0, 0), thickness=2)
# Print bbox corners.
img_points = undist.project_points(obs.get_bbox().transpose())
for img_point in img_points:
cv2.circle(undistorted, (int(img_point[0]), int(img_point[1])), radius = 5, color = (0, 255, 0), thickness=2)
cv2.imwrite('/data/dev/camera/undist_{}.png'.format(count), undistorted)
output_count += 1
count += 1
if desired_count is not None and output_count == desired_count:
return
def generate_camera_images(bag_file, mc, tracklet_file):
camera_converter = cc.CameraConverter()
im = None
generator = ns.generate_numpystream(bag_file, tracklet_file)
for numpydata in generator:
im = numpydata.image
obs = numpydata.obs
if im is not None:
im = sd.undistort_and_crop(im, camera_converter, mc)
# Must yield item for each frame in generator.
yield im, obs
def __init__(self, demo_net):
assert demo_net == 'squeezeDet' or demo_net == 'squeezeDet+' \
or demo_net == 'didi', \
'Selected neural net architecture not supported: {}'.format(demo_net)
self.demo_net = demo_net
self.mc = None
self.model = None
self.sess = None
self._prepare_graph()
self.camera_converter = cc.CameraConverter()
def try_augmenting_camera_boxes():
camera_converter = cc.CameraConverter()
bagdir = '/data/bags/didi-round2/release/car/training/suburu_leading_front_left'
bt = mb.find_bag_tracklets(bagdir, '/data/tracklets')
generator = generate_training_data_multi(bt)
count = 0
for bbox, label, im in generator:
new_bbox, new_label = augment_example(bbox, label, camera_converter)
print('bbox', bbox)
print('new_bbox', new_bbox)
print('label', label)
print('new_label', new_label)
count += 1
if count == 10:
return
def generate_camera_boxes_dir(train_dir, index_file, augment, infinite = True):
camera_converter = cc.CameraConverter()
while infinite:
ids = []
with open(os.path.join(train_dir, index_file), 'r') as f:
for id in f:
ids.append(int(id))
bboxdir, labeldir, imagedir = get_bbox_label_dirs(train_dir)
for id in ids:
bbox_path = get_bbox_path(bboxdir, id)
bbox = np.loadtxt(bbox_path)
label_path = get_label_path(labeldir, id)
label = np.loadtxt(label_path)
if augment:
(bbox, label) = augment_example(bbox, label, camera_converter)
yield bbox, label
def generate_training_data(bag_file, tracklet_file):
mc = sd.get_model_config(demo_net = 'squeezeDet')
camera_converter = cc.CameraConverter()
generator = generate_top_boxes(bag_file, tracklet_file)
for top_car, top_ped, (frame_idx, obs), im in generator:
top_obs = None
if obs.object_type == 'Car' and top_car is not None:
top_obs = top_car
elif obs.object_type == 'Pedestrian' and top_ped is not None:
top_obs = top_ped
if top_obs is not None:
(box, prob, class_idx) = top_obs
if (box is not None and
camera_converter.obstacle_is_in_view(obs)):
yield (np.array([box[0], box[1], box[2], box[3], prob, class_idx]),
np.array([obs.position[0], obs.position[1], obs.position[2], obs.yaw]),
im)
def __init__(self):
self.squeezedet = sd.SqueezeDetector(demo_net = 'squeezeDet')
self.box_to_pose_predictor = get_latest_predictor()
self.camera_converter = cc.CameraConverter()