def load_samples(self, indices):
""" Loads input-output data for a set of samples. Should only be
called when a particular sample dict is required. Otherwise,
samples should be provided by the next_batch function
Args:
indices: A list of sample indices from the dataset.sample_list
to be loaded
Return:
samples: a list of data sample dicts
"""
sample_dicts = []
for sample_idx in indices:
sample = self.sample_list[sample_idx]
sample_name = sample.name
# Only read labels if they exist
if self.has_labels:
# Read mini batch first to see if it is empty
anchors_info = self.get_anchors_info(sample_name)
if (not anchors_info) and self.train_val_test == 'train' \
and (not self.train_on_all_samples):
empty_sample_dict = {
constants.KEY_SAMPLE_NAME: sample_name,
constants.KEY_ANCHORS_INFO: anchors_info
}
return [empty_sample_dict]
obj_labels = obj_utils.read_labels(self.label_dir,
int(sample_name))
# Only use objects that match dataset classes
obj_labels = self.kitti_utils.filter_labels(obj_labels)
else:
obj_labels = None
anchors_info = []
label_anchors = np.zeros((1, 6))
label_boxes_3d = np.zeros((1, 7))
label_classes = np.zeros(1)
img_idx = int(sample_name)
# Load image (BGR -> RGB)
cv_bgr_image = cv2.imread(self.get_rgb_image_path(sample_name))
rgb_image = cv_bgr_image[..., ::-1]
image_shape = rgb_image.shape[0:2]
image_input = rgb_image
# Get ground plane
ground_plane = obj_utils.get_road_plane(int(sample_name),
self.planes_dir)
# Get calibration
stereo_calib_p2 = calib_utils.read_calibration(
self.calib_dir, int(sample_name)).p2
point_cloud = self.kitti_utils.get_point_cloud(
self.bev_source, img_idx, image_shape)
# Augmentation (Flipping)
if kitti_aug.AUG_FLIPPING in sample.augs:
image_input = kitti_aug.flip_image(image_input)
point_cloud = kitti_aug.flip_point_cloud(point_cloud)
obj_labels = [
kitti_aug.flip_label_in_3d_only(obj) for obj in obj_labels
]
ground_plane = kitti_aug.flip_ground_plane(ground_plane)
stereo_calib_p2 = kitti_aug.flip_stereo_calib_p2(
stereo_calib_p2, image_shape)
# Augmentation (Image Jitter)
if kitti_aug.AUG_PCA_JITTER in sample.augs:
image_input[:, :,
0:3] = kitti_aug.apply_pca_jitter(image_input[:, :,
0:3])
if obj_labels is not None:
label_boxes_3d = np.asarray([
box_3d_encoder.object_label_to_box_3d(obj_label)
for obj_label in obj_labels
])
label_classes = [
self.kitti_utils.class_str_to_index(obj_label.type)
for obj_label in obj_labels
]
label_classes = np.asarray(label_classes, dtype=np.int32)
# Return empty anchors_info if no ground truth after filtering
if len(label_boxes_3d) == 0:
anchors_info = []
if self.train_on_all_samples:
# If training without any positive labels, we cannot
# set these to zeros, because later on the offset calc
# uses log on these anchors. So setting any arbitrary
# number here that does not break the offset calculation
# should work, since the negative samples won't be
# regressed in any case.
dummy_anchors = [[-1000, -1000, -1000, 1, 1, 1]]
label_anchors = np.asarray(dummy_anchors)
dummy_boxes = [[-1000, -1000, -1000, 1, 1, 1, 0]]
label_boxes_3d = np.asarray(dummy_boxes)
else:
label_anchors = np.zeros((1, 6))
label_boxes_3d = np.zeros((1, 7))
label_classes = np.zeros(1)
else:
label_anchors = box_3d_encoder.box_3d_to_anchor(
label_boxes_3d, ortho_rotate=True)
# Create BEV maps
bev_images = self.kitti_utils.create_bev_maps(
point_cloud, ground_plane)
height_maps = bev_images.get('height_maps')
density_map = bev_images.get('density_map')
bev_input = np.dstack((*height_maps, density_map))
sample_dict = {
constants.KEY_LABEL_BOXES_3D: label_boxes_3d,
constants.KEY_LABEL_ANCHORS: label_anchors,
constants.KEY_LABEL_CLASSES: label_classes,
constants.KEY_IMAGE_INPUT: image_input,
constants.KEY_BEV_INPUT: bev_input,
constants.KEY_ANCHORS_INFO: anchors_info,
constants.KEY_POINT_CLOUD: point_cloud,
constants.KEY_GROUND_PLANE: ground_plane,
constants.KEY_STEREO_CALIB_P2: stereo_calib_p2,
constants.KEY_SAMPLE_NAME: sample_name,
constants.KEY_SAMPLE_AUGS: sample.augs
}
sample_dicts.append(sample_dict)
return sample_dicts
def main():
"""
Displays the bird's eye view maps for a KITTI sample.
"""
##############################
# Options
##############################
bev_generator = 'slices'
slices_config = \
"""
slices {
height_lo: -0.2
height_hi: 2.3
num_slices: 5
}
"""
# Use None for a random image
img_idx = None
# img_idx = 142
# img_idx = 191
show_ground_truth = True # Whether to overlay ground_truth boxes
point_cloud_source = 'lidar'
##############################
# End of Options
##############################
dataset_config = DatasetBuilder.copy_config(DatasetBuilder.KITTI_VAL)
dataset_config = DatasetBuilder.merge_defaults(dataset_config)
# Overwrite bev_generator
if bev_generator == 'slices':
text_format.Merge(slices_config,
dataset_config.kitti_utils_config.bev_generator)
else:
raise ValueError('Invalid bev_generator')
dataset = DatasetBuilder.build_kitti_dataset(dataset_config,
use_defaults=False)
if img_idx is None:
img_idx = int(random.random() * dataset.num_samples)
sample_name = "{:06}".format(img_idx)
print('=== Showing BEV maps for image: {}.png ==='.format(sample_name))
# Load image
image = cv2.imread(dataset.get_rgb_image_path(sample_name))
image_shape = image.shape[0:2]
kitti_utils = dataset.kitti_utils
print(image_shape)
point_cloud = kitti_utils.get_point_cloud(point_cloud_source, img_idx,
image_shape)
ground_plane = kitti_utils.get_ground_plane(sample_name)
bev_images = kitti_utils.create_bev_maps(point_cloud, ground_plane)
height_maps = np.array(bev_images.get("height_maps"))
density_map = np.array(bev_images.get("density_map"))
box_points, box_points_norm = [None, None]
if show_ground_truth:
# Get projected boxes
obj_labels = obj_utils.read_labels(dataset.label_dir, img_idx)
print(obj_labels)
filtered_objs = obj_labels
label_boxes = []
for label in filtered_objs:
box = box_3d_encoder.object_label_to_box_3d(label)
label_boxes.append(box)
label_boxes = np.array(label_boxes)
print(label_boxes)
box_points, box_points_norm = box_3d_projector.project_to_bev(
label_boxes, [[-40, 40], [0, 70]])
rgb_img_size = (np.array((1242, 375)) * 0.75).astype(np.int16)
img_x_start = 60
img_y_start = 330
img_x = img_x_start
img_y = img_y_start
img_w = 400
img_h = 350
img_titlebar_h = 20
# Show images
vis_utils.cv2_show_image("Image",
image,
size_wh=rgb_img_size,
location_xy=(img_x, 0))
# Height maps
for map_idx in range(len(height_maps)):
height_map = height_maps[map_idx]
height_map = draw_boxes(height_map, box_points_norm)
vis_utils.cv2_show_image("Height Map {}".format(map_idx),
height_map,
size_wh=(img_w, img_h),
location_xy=(img_x, img_y))
img_x += img_w
# Wrap around
if (img_x + img_w) > 1920:
img_x = img_x_start
img_y += img_h + img_titlebar_h
# Density map
density_map = draw_boxes(density_map, box_points_norm)
vis_utils.cv2_show_image("Density Map",
density_map,
size_wh=(img_w, img_h),
location_xy=(img_x, img_y))
cv2.waitKey()
def get_clusters(self):
"""
Calculates clusters for each class
Returns:
all_clusters: list of clusters for each class
all_std_devs: list of cluster standard deviations for each class
"""
classes = self._dataset.classes
num_clusters = self._dataset.num_clusters
all_clusters = [[] for _ in range(len(classes))]
all_std_devs = [[] for _ in range(len(classes))]
classes_not_loaded = []
# Try to read from file first
for class_idx in range(len(classes)):
clusters, std_devs = self._read_clusters_from_file(
self._dataset, classes[class_idx], num_clusters[class_idx])
if clusters is not None:
all_clusters[class_idx].extend(np.asarray(clusters))
all_std_devs[class_idx].extend(np.asarray(std_devs))
else:
classes_not_loaded.append(class_idx)
# Return the data flattened into N x 3 arrays
if len(classes_not_loaded) == 0:
return all_clusters, all_std_devs
# Calculate the remaining clusters
# Load labels corresponding to the sample list for clustering
sample_list = self._dataset.load_sample_names(self.cluster_split)
all_labels = [[] for _ in range(len(classes))]
num_samples = len(sample_list)
for sample_idx in range(num_samples):
sys.stdout.write("\rClustering labels {} / {}".format(
sample_idx + 1, num_samples))
sys.stdout.flush()
sample_name = sample_list[sample_idx]
img_idx = int(sample_name)
obj_labels = obj_utils.read_labels(self._dataset.label_dir,
img_idx)
filtered_labels = LabelClusterUtils._filter_labels_by_class(
obj_labels, self._dataset.classes)
for class_idx in range(len(classes)):
all_labels[class_idx].extend(filtered_labels[class_idx])
print("\nFinished reading labels, clustering data...\n")
# Cluster
for class_idx in classes_not_loaded:
labels_for_class = np.array(all_labels[class_idx])
n_clusters_for_class = num_clusters[class_idx]
if len(labels_for_class) < n_clusters_for_class:
raise ValueError(
"Number of samples is less than number of clusters "
"{} < {}".format(len(labels_for_class),
n_clusters_for_class))
k_means = KMeans(n_clusters=n_clusters_for_class,
random_state=0).fit(labels_for_class)
clusters_for_class = []
std_devs_for_class = []
for cluster_idx in range(len(k_means.cluster_centers_)):
cluster_centre = k_means.cluster_centers_[cluster_idx]
labels_in_cluster = labels_for_class[
k_means.labels_ == cluster_idx]
# Calculate std. dev
std_dev = np.std(labels_in_cluster, axis=0)
formatted_cluster = [float('%.3f' % value)
for value in cluster_centre]
formatted_std_dev = [float('%.3f' % value)
for value in std_dev]
clusters_for_class.append(formatted_cluster)
std_devs_for_class.append(formatted_std_dev)
# Write to files
file_path = self._get_cluster_file_path(self._dataset,
classes[class_idx],
num_clusters[class_idx])
self._write_clusters_to_file(file_path, clusters_for_class,
std_devs_for_class)
# Add to full list
all_clusters[class_idx].extend(np.asarray(clusters_for_class))
all_std_devs[class_idx].extend(np.asarray(std_devs_for_class))
# Return the data flattened into N x 3 arrays
return all_clusters, all_std_devs
def preprocess(self, indices):
"""Preprocesses anchor info and saves info to files
Args:
indices (int array): sample indices to process.
If None, processes all samples
"""
# Get anchor stride for class
anchor_strides = self._anchor_strides
dataset = self._dataset
dataset_utils = self._dataset.kitti_utils
classes_name = dataset.classes_name
# Make folder if it doesn't exist yet
output_dir = self.mini_batch_utils.get_file_path(classes_name,
anchor_strides,
sample_name=None)
os.makedirs(output_dir, exist_ok=True)
# Get clusters for class
all_clusters_sizes, _ = dataset.get_cluster_info()
anchor_generator = grid_anchor_3d_generator.GridAnchor3dGenerator()
# Load indices of data_split
all_samples = dataset.sample_list
if indices is None:
indices = np.arange(len(all_samples))
num_samples = len(indices)
# For each image in the dataset, save info on the anchors
for sample_idx in indices:
# Get image name for given cluster
sample_name = all_samples[sample_idx].name
img_idx = int(sample_name)
# Check for existing files and skip to the next
if self._check_for_existing(classes_name, anchor_strides,
sample_name):
print("{} / {}: Sample already preprocessed".format(
sample_idx + 1, num_samples, sample_name))
continue
# Get ground truth and filter based on difficulty
ground_truth_list = obj_utils.read_labels(dataset.label_dir,
img_idx)
# Filter objects to dataset classes
filtered_gt_list = dataset_utils.filter_labels(ground_truth_list)
filtered_gt_list = np.asarray(filtered_gt_list)
# Filtering by class has no valid ground truth, skip this image
if len(filtered_gt_list) == 0:
print("{} / {} No {}s for sample {} "
"(Ground Truth Filter)".format(sample_idx + 1,
num_samples, classes_name,
sample_name))
# Output an empty file and move on to the next image.
self._save_to_file(classes_name, anchor_strides, sample_name)
continue
# Get ground plane
ground_plane = obj_utils.get_road_plane(img_idx,
dataset.planes_dir)
image = Image.open(dataset.get_rgb_image_path(sample_name))
image_shape = [image.size[1], image.size[0]]
# Generate sliced 2D voxel grid for filtering
vx_grid_2d = dataset_utils.create_sliced_voxel_grid_2d(
sample_name,
source=dataset.bev_source,
image_shape=image_shape)
# List for merging all anchors
all_anchor_boxes_3d = []
# Create anchors for each class
for class_idx in range(len(dataset.classes)):
# Generate anchors for all classes
grid_anchor_boxes_3d = anchor_generator.generate(
area_3d=self._area_extents,
anchor_3d_sizes=all_clusters_sizes[class_idx],
anchor_stride=self._anchor_strides[class_idx],
ground_plane=ground_plane)
all_anchor_boxes_3d.extend(grid_anchor_boxes_3d)
# Filter empty anchors
all_anchor_boxes_3d = np.asarray(all_anchor_boxes_3d)
anchors = box_3d_encoder.box_3d_to_anchor(all_anchor_boxes_3d)
empty_anchor_filter = anchor_filter.get_empty_anchor_filter_2d(
anchors, vx_grid_2d, self._density_threshold)
# Calculate anchor info
anchors_info = self._calculate_anchors_info(
all_anchor_boxes_3d, empty_anchor_filter, filtered_gt_list)
anchor_ious = anchors_info[:, self.mini_batch_utils.col_ious]
valid_iou_indices = np.where(anchor_ious > 0.0)[0]
print("{} / {}:"
"{:>6} anchors, "
"{:>6} iou > 0.0, "
"for {:>3} {}(s) for sample {}".format(
sample_idx + 1, num_samples, len(anchors_info),
len(valid_iou_indices), len(filtered_gt_list),
classes_name, sample_name))
# Save anchors info
self._save_to_file(classes_name, anchor_strides, sample_name,
anchors_info)
def main():
"""This demo shows RPN proposals and AVOD predictions in 3D
and 2D in image space. Given certain thresholds for proposals
and predictions, it selects and draws the bounding boxes on
the image sample. It goes through the entire proposal and
prediction samples for the given dataset split.
The proposals, overlaid, and prediction images can be toggled on or off
separately in the options section.
The prediction score and IoU with ground truth can be toggled on or off
as well, shown as (score, IoU) above the detection.
"""
dataset_config = DatasetBuilder.copy_config(DatasetBuilder.KITTI_VAL)
##############################
# Options
##############################
dataset_config.data_split = 'val'
fig_size = (10, 6.1)
rpn_score_threshold = 0.1
avod_score_threshold = 0.1
gt_classes = ['Car']
# gt_classes = ['Pedestrian', 'Cyclist']
# Overwrite this to select a specific checkpoint
global_step = None
checkpoint_name = 'avod_cars_example'
# Drawing Toggles
draw_proposals_separate = False
draw_overlaid = False
draw_predictions_separate = True
# Show orientation for both GT and proposals/predictions
draw_orientations_on_prop = False
draw_orientations_on_pred = False
# Draw 2D bounding boxes
draw_projected_2d_boxes = True
# Save images for samples with no detections
save_empty_images = True
draw_score = True
draw_iou = True
##############################
# End of Options
##############################
# Get the dataset
dataset = DatasetBuilder.build_kitti_dataset(dataset_config)
# Setup Paths
predictions_dir = avod.root_dir() + \
'/data/outputs/' + checkpoint_name + '/predictions'
proposals_and_scores_dir = predictions_dir + \
'/proposals_and_scores/' + dataset.data_split
predictions_and_scores_dir = predictions_dir + \
'/final_predictions_and_scores/' + dataset.data_split
# Output images directories
output_dir_base = predictions_dir + '/images_2d'
# Get checkpoint step
steps = os.listdir(proposals_and_scores_dir)
steps.sort(key=int)
print('Available steps: {}'.format(steps))
# Use latest checkpoint if no index provided
if global_step is None:
global_step = steps[-1]
if draw_proposals_separate:
prop_out_dir = output_dir_base + '/proposals/{}/{}/{}'.format(
dataset.data_split, global_step, rpn_score_threshold)
if not os.path.exists(prop_out_dir):
os.makedirs(prop_out_dir)
print('Proposal images saved to:', prop_out_dir)
if draw_overlaid:
overlaid_out_dir = output_dir_base + '/overlaid/{}/{}/{}'.format(
dataset.data_split, global_step, avod_score_threshold)
if not os.path.exists(overlaid_out_dir):
os.makedirs(overlaid_out_dir)
print('Overlaid images saved to:', overlaid_out_dir)
if draw_predictions_separate:
pred_out_dir = output_dir_base + '/predictions/{}/{}/{}'.format(
dataset.data_split, global_step,
avod_score_threshold)
if not os.path.exists(pred_out_dir):
os.makedirs(pred_out_dir)
print('Prediction images saved to:', pred_out_dir)
# Rolling average array of times for time estimation
avg_time_arr_length = 10
last_times = np.repeat(time.time(), avg_time_arr_length) + \
np.arange(avg_time_arr_length)
for sample_idx in range(dataset.num_samples):
# Estimate time remaining with 5 slowest times
start_time = time.time()
last_times = np.roll(last_times, -1)
last_times[-1] = start_time
avg_time = np.mean(np.sort(np.diff(last_times))[-5:])
samples_remaining = dataset.num_samples - sample_idx
est_time_left = avg_time * samples_remaining
# Print progress and time remaining estimate
sys.stdout.write('\rSaving {} / {}, Avg Time: {:.3f}s, '
'Time Remaining: {:.2f}s'. format(
sample_idx + 1,
dataset.num_samples,
avg_time,
est_time_left))
sys.stdout.flush()
sample_name = dataset.sample_names[sample_idx]
img_idx = int(sample_name)
##############################
# Proposals
##############################
if draw_proposals_separate or draw_overlaid:
# Load proposals from files
proposals_file_path = proposals_and_scores_dir + \
"/{}/{}.txt".format(global_step, sample_name)
if not os.path.exists(proposals_file_path):
print('Sample {}: No proposals, skipping'.format(sample_name))
continue
print('Sample {}: Drawing proposals'.format(sample_name))
proposals_and_scores = np.loadtxt(proposals_file_path)
proposal_boxes_3d = proposals_and_scores[:, 0:7]
proposal_scores = proposals_and_scores[:, 7]
# Apply score mask to proposals
score_mask = proposal_scores > rpn_score_threshold
proposal_boxes_3d = proposal_boxes_3d[score_mask]
proposal_scores = proposal_scores[score_mask]
proposal_objs = \
[box_3d_encoder.box_3d_to_object_label(proposal,
obj_type='Proposal')
for proposal in proposal_boxes_3d]
##############################
# Predictions
##############################
if draw_predictions_separate or draw_overlaid:
predictions_file_path = predictions_and_scores_dir + \
"/{}/{}.txt".format(global_step,
sample_name)
if not os.path.exists(predictions_file_path):
continue
# Load predictions from files
predictions_and_scores = np.loadtxt(
predictions_and_scores_dir +
"/{}/{}.txt".format(global_step,
sample_name))
print(predictions_and_scores_dir)
prediction_boxes_3d = predictions_and_scores[:, 0:7]
prediction_scores = predictions_and_scores[:, 7]
prediction_class_indices = predictions_and_scores[:, 8]
# process predictions only if we have any predictions left after
# masking
if len(prediction_boxes_3d) > 0:
# Apply score mask
avod_score_mask = prediction_scores >= avod_score_threshold
prediction_boxes_3d = prediction_boxes_3d[avod_score_mask]
prediction_scores = prediction_scores[avod_score_mask]
prediction_class_indices = \
prediction_class_indices[avod_score_mask]
# # Swap l, w for predictions where w > l
# swapped_indices = \
# prediction_boxes_3d[:, 4] > prediction_boxes_3d[:, 3]
# prediction_boxes_3d = np.copy(prediction_boxes_3d)
# prediction_boxes_3d[swapped_indices, 3] = \
# prediction_boxes_3d[swapped_indices, 4]
# prediction_boxes_3d[swapped_indices, 4] = \
# prediction_boxes_3d[swapped_indices, 3]
##############################
# Ground Truth
##############################
# Get ground truth labels
if dataset.has_labels:
gt_objects = obj_utils.read_labels(dataset.label_dir, img_idx)
else:
gt_objects = []
# Filter objects to desired difficulty
filtered_gt_objs = dataset.kitti_utils.filter_labels(
gt_objects, classes=gt_classes)
boxes2d, _, _ = obj_utils.build_bbs_from_objects(
filtered_gt_objs, class_needed=gt_classes)
image_path = dataset.get_rgb_image_path(sample_name)
image = Image.open(image_path)
image_size = image.size
# Read the stereo calibration matrix for visualization
stereo_calib = calib_utils.read_calibration(dataset.calib_dir,
img_idx)
calib_p2 = stereo_calib.p2
##############################
# Reformat and prepare to draw
##############################
if draw_proposals_separate or draw_overlaid:
proposals_as_anchors = box_3d_encoder.box_3d_to_anchor(
proposal_boxes_3d)
proposal_boxes, _ = anchor_projector.project_to_image_space(
proposals_as_anchors, calib_p2, image_size)
num_of_proposals = proposal_boxes_3d.shape[0]
prop_fig, prop_2d_axes, prop_3d_axes = \
vis_utils.visualization(dataset.rgb_image_dir,
img_idx,
display=False)
draw_proposals(filtered_gt_objs,
calib_p2,
num_of_proposals,
proposal_objs,
proposal_boxes,
prop_2d_axes,
prop_3d_axes,
draw_orientations_on_prop)
if draw_proposals_separate:
# Save just the proposals
filename = prop_out_dir + '/' + sample_name + '.png'
plt.savefig(filename)
if not draw_overlaid:
plt.close(prop_fig)
if draw_overlaid or draw_predictions_separate:
if len(prediction_boxes_3d) > 0:
# Project the 3D box predictions to image space
image_filter = []
final_boxes_2d = []
for i in range(len(prediction_boxes_3d)):
box_3d = prediction_boxes_3d[i, 0:7]
img_box = box_3d_projector.project_to_image_space(
box_3d, calib_p2,
truncate=True, image_size=image_size,
discard_before_truncation=False)
if img_box is not None:
image_filter.append(True)
final_boxes_2d.append(img_box)
else:
image_filter.append(False)
final_boxes_2d = np.asarray(final_boxes_2d)
final_prediction_boxes_3d = prediction_boxes_3d[image_filter]
final_scores = prediction_scores[image_filter]
final_class_indices = prediction_class_indices[image_filter]
num_of_predictions = final_boxes_2d.shape[0]
# Convert to objs
final_prediction_objs = \
[box_3d_encoder.box_3d_to_object_label(
prediction, obj_type='Prediction')
for prediction in final_prediction_boxes_3d]
for (obj, score) in zip(final_prediction_objs, final_scores):
obj.score = score
else:
if save_empty_images:
pred_fig, pred_2d_axes, pred_3d_axes = \
vis_utils.visualization(dataset.rgb_image_dir,
img_idx,
display=False,
fig_size=fig_size)
filename = pred_out_dir + '/' + sample_name + '.png'
plt.savefig(filename)
plt.close(pred_fig)
continue
if draw_overlaid:
# Overlay prediction boxes on image
draw_predictions(filtered_gt_objs,
calib_p2,
num_of_predictions,
final_prediction_objs,
final_class_indices,
final_boxes_2d,
prop_2d_axes,
prop_3d_axes,
draw_score,
draw_iou,
gt_classes,
draw_orientations_on_pred)
filename = overlaid_out_dir + '/' + sample_name + '.png'
plt.savefig(filename)
plt.close(prop_fig)
if draw_predictions_separate:
# Now only draw prediction boxes on images
# on a new figure handler
if draw_projected_2d_boxes:
pred_fig, pred_2d_axes, pred_3d_axes = \
vis_utils.visualization(dataset.rgb_image_dir,
img_idx,
display=False,
fig_size=fig_size)
draw_predictions(filtered_gt_objs,
calib_p2,
num_of_predictions,
final_prediction_objs,
final_class_indices,
final_boxes_2d,
pred_2d_axes,
pred_3d_axes,
draw_score,
draw_iou,
gt_classes,
draw_orientations_on_pred)
else:
pred_fig, pred_3d_axes = \
vis_utils.visualize_single_plot(
dataset.rgb_image_dir, img_idx, display=False)
draw_3d_predictions(filtered_gt_objs,
calib_p2,
num_of_predictions,
final_prediction_objs,
final_class_indices,
final_boxes_2d,
pred_3d_axes,
draw_score,
draw_iou,
gt_classes,
draw_orientations_on_pred)
filename = pred_out_dir + '/' + sample_name + '.png'
plt.savefig(filename)
plt.close(pred_fig)
print('\nDone')
def test(model_config, eval_config, dataset_config, data_split, ckpt_indices):
# Overwrite the defaults
dataset_config = config_builder.proto_to_obj(dataset_config)
dataset_config.data_split = data_split
dataset_config.data_split_dir = 'training'
if data_split == 'test':
dataset_config.data_split_dir = 'testing'
eval_config.eval_mode = 'test'
eval_config.evaluate_repeatedly = False
dataset_config.has_labels = False
# Enable this to see the actually memory being used
eval_config.allow_gpu_mem_growth = True
eval_config = config_builder.proto_to_obj(eval_config)
# Grab the checkpoint indices to evaluate
eval_config.ckpt_indices = ckpt_indices
# Remove augmentation during evaluation in test mode
dataset_config.aug_list = []
# Build the dataset object
dataset = DatasetBuilder.build_kitti_dataset(dataset_config,
use_defaults=False)
# Setup the model
model_name = model_config.model_name
# Overwrite repeated field
model_config = config_builder.proto_to_obj(model_config)
# Switch path drop off during evaluation
model_config.path_drop_probabilities = [1.0, 1.0]
with tf.Graph().as_default():
if model_name == 'avod_model':
model = AvodModel(model_config,
train_val_test=eval_config.eval_mode,
dataset=dataset)
elif model_name == 'rpn_model':
model = RpnModel(model_config,
train_val_test=eval_config.eval_mode,
dataset=dataset)
else:
raise ValueError('Invalid model name {}'.format(model_name))
#model_evaluator = Evaluator(model, dataset_config, eval_config)
#model_evaluator.run_latest_checkpoints()
# Create a variable tensor to hold the global step
global_step_tensor = tf.Variable(0,
trainable=False,
name='global_step')
allow_gpu_mem_growth = eval_config.allow_gpu_mem_growth
if allow_gpu_mem_growth:
# GPU memory config
config = tf.ConfigProto()
config.gpu_options.allow_growth = allow_gpu_mem_growth
_sess = tf.Session(config=config)
else:
_sess = tf.Session()
_prediction_dict = model.build()
_saver = tf.train.Saver()
trainer_utils.load_checkpoints(
model_config.paths_config.checkpoint_dir, _saver)
num_checkpoints = len(_saver.last_checkpoints)
print("test:", num_checkpoints)
checkpoint_to_restore = _saver.last_checkpoints[num_checkpoints - 1]
_saver.restore(_sess, checkpoint_to_restore)
num_samples = model.dataset.num_samples
num_valid_samples = 0
current_epoch = model.dataset.epochs_completed
data_dir = '/media/peiliang/Data/dataset/kitti/raw_data/2011_10_03/raw_tracking_training_image_02_0020'
for image_idx in range(0, 100000):
IMAGE_PATH = os.path.join(
data_dir, 'image_02/data/{:010d}.png'.format(image_idx))
LIDAR_PATH = os.path.join(
data_dir, 'velodyne_points/data/{:010d}.bin'.format(image_idx))
CALIB_DIR = '/media/peiliang/Data/dataset/kitti/raw_data/2011_10_03'
# Keep track of feed_dict speed
start_time = time.time()
feed_dict = model._rpn_model.create_single_feed_dict(
IMAGE_PATH, LIDAR_PATH, CALIB_DIR)
feed_dict_time = time.time() - start_time
print("tracking mode")
inference_start_time = time.time()
# Don't calculate loss or run summaries for test
predictions = _sess.run(_prediction_dict, feed_dict=feed_dict)
inference_time = time.time() - inference_start_time
print("inference time:", inference_time)
predictions_and_scores = get_avod_predicted_boxes_3d_and_scores(
predictions)
#print(predictions_and_scores)
#im_path = os.path.join(dataset_dir, 'training/image_2/{:06d}.png'.format(img_idx))
#im = cv2.imread(im_path)
#cv2.imshow('result',im)
#cv2.waitKey(30)
prediction_boxes_3d = predictions_and_scores[:, 0:7]
prediction_scores = predictions_and_scores[:, 7]
prediction_class_indices = predictions_and_scores[:, 8]
gt_classes = ['Car']
fig_size = (10, 6.1)
avod_score_threshold = 0.1
if len(prediction_boxes_3d) > 0:
# Apply score mask
avod_score_mask = prediction_scores >= avod_score_threshold
prediction_boxes_3d = prediction_boxes_3d[avod_score_mask]
prediction_scores = prediction_scores[avod_score_mask]
prediction_class_indices = \
prediction_class_indices[avod_score_mask]
if len(prediction_boxes_3d) > 0:
dataset_dir = model.dataset.dataset_dir
sample_name = (
model.dataset.sample_names[model.dataset._index_in_epoch -
1])
img_idx = int(sample_name)
print("frame_index", image_idx)
image = Image.open(IMAGE_PATH)
image_size = image.size
if model.dataset.has_labels:
gt_objects = obj_utils.read_labels(dataset.label_dir,
img_idx)
else:
gt_objects = []
filtered_gt_objs = model.dataset.kitti_utils.filter_labels(
gt_objects, classes=gt_classes)
stereo_calib = calib_utils.read_raw_calibration(CALIB_DIR)
stereo_calib_p2 = stereo_calib.p2
# Project the 3D box predictions to image space
image_filter = []
final_boxes_2d = []
for i in range(len(prediction_boxes_3d)):
box_3d = prediction_boxes_3d[i, 0:7]
img_box = box_3d_projector.project_to_image_space(
box_3d,
stereo_calib_p2,
truncate=True,
image_size=image_size,
discard_before_truncation=False)
if img_box is not None:
image_filter.append(True)
final_boxes_2d.append(img_box)
else:
image_filter.append(False)
final_boxes_2d = np.asarray(final_boxes_2d)
final_prediction_boxes_3d = prediction_boxes_3d[image_filter]
final_scores = prediction_scores[image_filter]
final_class_indices = prediction_class_indices[image_filter]
num_of_predictions = final_boxes_2d.shape[0]
# Convert to objs
final_prediction_objs = \
[box_3d_encoder.box_3d_to_object_label(
prediction, obj_type='Prediction')
for prediction in final_prediction_boxes_3d]
for (obj, score) in zip(final_prediction_objs, final_scores):
obj.score = score
pred_fig, pred_2d_axes, pred_3d_axes = \
vis_utils.visualization_from_file(IMAGE_PATH,
display=False,
fig_size=fig_size)
draw_predictions(filtered_gt_objs, stereo_calib_p2,
num_of_predictions, final_prediction_objs,
final_class_indices, final_boxes_2d,
pred_2d_axes, pred_3d_axes, True, True,
gt_classes, False)
#cv2.imshow('result',pred_fig)
print(type(pred_fig))
pred_fig.canvas.draw()
img = np.fromstring(pred_fig.canvas.tostring_rgb(),
dtype=np.uint8,
sep='')
img = img.reshape(pred_fig.canvas.get_width_height()[::-1] +
(3, ))
cv2.imshow('result', img)
#draw bird view
kitti_utils = model.dataset.kitti_utils
ground_plane = obj_utils.get_road_plane_from_file(CALIB_DIR)
# Get calibration
point_cloud = kitti_utils.get_point_cloud_from_file(
'lidar', stereo_calib, LIDAR_PATH, (370, 1242))
bev_images = kitti_utils.create_bev_maps(
point_cloud, ground_plane)
density_map = np.array(bev_images.get("density_map"))
_, box_points_norm = box_3d_projector.project_to_bev(
final_prediction_boxes_3d, [[-40, 40], [0, 70]])
density_map = draw_boxes(density_map, box_points_norm)
cv2.imshow('lidar', density_map)
cv2.waitKey(-1)