def do_sampling_step(self, ballness, new_time):
"""
Importance sampling step of Bootstrap Filter.
See http://www.cs.ubc.ca/%7Earnaud/doucet_defreitas_gordon_smcbookintro.ps (page 11).
"""
likelihood = numpy.exp(ballness)
(w, h) = likelihood.shape
transform = numpy.dot(transformation.rectangle_to_pixels(w, h), transformation.scale(1.0/metrics.FIELD_WIDTH, 1.0/metrics.FIELD_HEIGHT))
for particle in self.particles:
# Evolve particles
if self.time is None:
timedelta = 0.0
else:
timedelta = new_time - self.time
particle.evolve(timedelta, self.model_settings)
# Update weights
particle.update_weight(likelihood, transform, self.model_settings)
print self.particles
# Normalize weights
weights_sum = sum([particle.weight for particle in self.particles])
for particle in self.particles:
particle.weight /= weights_sum
import cv2
from particlefilter import *
import transformation
import metrics
def generate_ballness(w, h):
return numpy.array([random.random() for i in xrange(w*h)]).reshape((w,h))
pf = ParticleFilter()
w = 500
h = 1000
time = 0.0
transform = numpy.dot(transformation.scale(metrics.FIELD_WIDTH, metrics.FIELD_HEIGHT), transformation.rectangle_to_pixels(w, h))
while True:
ballness = generate_ballness(w, h)
img = numpy.array([[[x,x,x] for x in row] for row in ballness]).reshape((w,h,3))
time += 0.5
n = pf.settings.num_particles
# Sampling step
pf.do_sampling_step(ballness, time)
for particle in pf.particles:
if particle.is_present():
point = tuple([int(x) for x in transformation.apply_projectivity(transform, particle.position)])
cv2.circle(img=img, center=point, radius=15, color=(2.0/n - particle.weight, 2.0/n - particle.weight, 1.0), thickness=3)
def radar_data_grapher_volumned(paths,
is_plot=False,
isCluster=False,
augmentation=(),
seeds=np.random.normal(0, 0.02, 5000),
isDataGen=True):
# utility directory to save the pyplots
radarData_path, videoData_path, mergedImg_path, out_path, identity_string = paths
radar_3dscatter_path = 'E:/indexPen/figures/utils/radar_3dscatter'
radar_data = list(pickle.load(open(radarData_path, 'rb')).items())
radar_data.sort(key=lambda x: x[0]) # sort by timestamp
videoData_list = os.listdir(videoData_path)
videoData_timestamps = list(
map(lambda x: float(x.strip('.jpg')), videoData_list))
style.use('fivethirtyeight')
white_color = 'rgb(255, 255, 255)'
black_color = 'rgb(0, 0, 0)'
red_color = 'rgb(255, 0, 0)'
DBSCAN_esp = 0.2
DBSCAN_minSamples = 3
# input data for the classifier that has the shape n*4*100, n being the number of samples
num_padding = 100
data_for_classifier = np.zeros((len(radar_data), num_padding, 4))
data_for_classifier_flattened = np.zeros(
(len(radar_data), 4 * num_padding + 1 + 1 +
1)) # + 1 + 1 for the timestamp as integer ratio
fnt = ImageFont.truetype("arial.ttf", 16)
# Retrieve the first timestamp
starting_timestamp = radar_data[0][0]
interval_index = 1
# removed and recreate the merged image folder
if is_plot:
if os.path.isdir(mergedImg_path):
shutil.rmtree(mergedImg_path)
os.mkdir(mergedImg_path)
volume_shape = (25, 25, 25)
interval_volume_list = []
volumes_for_this_interval = []
interval_sec = 5
sample_per_sec = 20
sample_per_interval = interval_sec * sample_per_sec
aug_string = ''
if augmentation:
print('Use augmentation: ' + str(augmentation))
for aug in augmentation:
aug_string += '_' + aug
else:
print('No augmentation applied')
print('Label Cheat-sheet:')
print('0 for A')
print('1 for D')
print('2 for L')
print('3 for M')
print('4 for P')
label_array = []
this_label = 0
for i, radarFrame in enumerate(radar_data):
# retrieve the data
timestamp, fData = radarFrame
# calculate the interval
if (timestamp - starting_timestamp) >= 5.0:
num_intervaled_samples = len(volumes_for_this_interval)
if num_intervaled_samples < sample_per_interval / 8:
raise Exception('Not Enough Data Points, killed')
# decide the label
inter_arg = 20
if interval_index % inter_arg == 1 or interval_index % inter_arg == 2:
this_label = 0
elif interval_index % inter_arg == 3 or interval_index % inter_arg == 4:
this_label = 1 # for label D
elif interval_index % inter_arg == 5 or interval_index % inter_arg == 6:
this_label = 2 # for label L
elif interval_index % inter_arg == 7 or interval_index % inter_arg == 8:
this_label = 3 # for label M
elif interval_index % inter_arg == 9 or interval_index % inter_arg == 10:
this_label = 4 # for label P
elif interval_index % inter_arg == 11 or interval_index % inter_arg == 12:
this_label = 5 # for label D
elif interval_index % inter_arg == 13 or interval_index % inter_arg == 14:
this_label = 6 # for label L
elif interval_index % inter_arg == 15 or interval_index % inter_arg == 16:
this_label = 7 # for label M
elif interval_index % inter_arg == 17 or interval_index % inter_arg == 18:
this_label = 8 # for label P
elif interval_index % inter_arg == 19 or interval_index % inter_arg == 0:
this_label = 9 # for label P
label_array.append(this_label) # for label A
print('Label for the last interval is ' + str(this_label) +
' Num Samples: ' + str(len(volumes_for_this_interval)))
print('')
# add padding, pre-padded
if len(volumes_for_this_interval) < sample_per_interval:
while len(volumes_for_this_interval) < sample_per_interval:
volumes_for_this_interval.insert(
0, np.expand_dims(np.zeros(volume_shape), axis=0))
elif len(
volumes_for_this_interval
) > sample_per_interval: # we take only the 75 most recent
volumes_for_this_interval = volumes_for_this_interval[-75:]
volumes_for_this_interval = np.asarray(volumes_for_this_interval)
interval_volume_list.append(volumes_for_this_interval)
volumes_for_this_interval = []
# increment the timestamp and interval index
starting_timestamp = starting_timestamp + 5.0
interval_index = interval_index + 1
# end of end of interval processing
print('Processing ' + str(i + 1) + ' of ' + str(len(radar_data)) +
', interval = ' + str(interval_index))
if is_plot:
mergedImg_path_intervaled = os.path.join(mergedImg_path,
str(interval_index - 1))
if not os.path.isdir(mergedImg_path_intervaled):
os.mkdir(mergedImg_path_intervaled)
closest_video_timestamp = min(videoData_timestamps,
key=lambda x: abs(x - timestamp))
closest_video_path = os.path.join(
videoData_path,
str(closest_video_timestamp) + '.jpg')
closest_video_img = Image.open(closest_video_path)
# plot the radar scatter
ax1 = plt.subplot(2, 2, 1, projection='3d')
ax1.set_xlim((-0.3, 0.3))
ax1.set_ylim((-0.3, 0.3))
ax1.set_zlim((-0.3, 0.3))
ax1.set_xlabel('X', fontsize=10)
ax1.set_ylabel('Y', fontsize=10)
ax1.set_zlabel('Z', fontsize=10)
ax1.set_title('Detected Points', fontsize=10)
# plot the detected points
ax1.scatter(fData['x'],
fData['y'],
fData['z'],
c=fData['doppler'],
marker='o')
data = np.asarray(
[fData['x'], fData['y'], fData['z'],
fData['doppler']]).transpose()
# Do DBSCAN cluster ###########################################
# map the points to their doppler value, this is for retrieving the doppler value after clustering
if isCluster:
doppler_dict = {}
for point in data:
doppler_dict[tuple(point[:3])] = point[3:]
# get rid of the doppler for clustering TODO should we consider the doppler in clustering?
data = data[:, :3]
db = DBSCAN(eps=DBSCAN_esp,
min_samples=DBSCAN_minSamples).fit(data)
core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
core_samples_mask[db.core_sample_indices_] = True
labels = db.labels_
# Number of clusters in labels, ignoring noise if present.
n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
n_noise_ = list(labels).count(-1)
if is_plot:
ax2 = plt.subplot(2, 2, 2, projection='3d')
ax2.set_xlim((-0.3, 0.3))
ax2.set_ylim((-0.3, 0.3))
ax2.set_zlim((-0.3, 0.3))
ax2.set_xlabel('X', fontsize=10)
ax2.set_ylabel('Y', fontsize=10)
ax2.set_zlabel('Z', fontsize=10)
ax2.set_title('Clustered Points', fontsize=10)
unique_labels = set(labels)
colors = [
plt.cm.Spectral(each)
for each in np.linspace(0, 1, len(unique_labels))
]
clusters = []
for k, col in zip(unique_labels, colors):
if k == -1:
# Black used for noise.
col = [0, 0, 0, 1]
class_member_mask = (labels == k)
xyz = data[class_member_mask & core_samples_mask]
if xyz.any(): # in case there are none objects
clusters.append(
xyz) # append this cluster data to the cluster list
# each cluster is a 3 * n matrix
xyz = data[class_member_mask & ~core_samples_mask]
if is_plot:
ax2.scatter(xyz[:, 0],
xyz[:, 1],
xyz[:, 2],
'o',
c=np.array([col]),
s=12,
marker='X') # plot the noise
# find the center for each cluster
clusters_centers = list(
map(
lambda xyz: np.array([
np.mean(xyz[:, 0]),
np.mean(xyz[:, 1]),
np.mean(xyz[:, 2])
]), clusters))
clusters.sort(
key=lambda xyz: distance.euclidean((0.0, 0.0, 0.0),
np.array([
np.mean(xyz[:, 0]),
np.mean(xyz[:, 1]),
np.mean(xyz[:, 2])
])))
# plot the clusters
for xyz, col in zip(clusters, colors):
if is_plot:
ax2.scatter(xyz[:, 0],
xyz[:, 1],
xyz[:, 2],
'o',
c=np.array([col]),
s=28,
marker='o') # plot the cluster points
#############################
# clear the hand cluster
hand_cluster = []
bbox = (0.2, 0.2, 0.2)
if len(clusters) > 0:
hand_cluster = clusters[0]
point_num = hand_cluster.shape[0]
# if the cluster is outside the 20*20*20 cm bounding box
distance_from_center = distance.euclidean(
(0.0, 0.0, 0.0),
np.array([
np.mean(hand_cluster[:, 0]),
np.mean(hand_cluster[:, 1]),
np.mean(hand_cluster[:, 2])
]))
if distance_from_center > distance.euclidean(
(0.0, 0.0, 0.0), bbox
): # if the core of the cluster is too far away from the center
hand_cluster = np.zeros(
(hand_cluster.shape[0], hand_cluster.shape[1] + 1))
else:
doppler_array = np.zeros((point_num, 1))
for j in range(point_num):
doppler_array[j:, ] = doppler_dict[tuple(
hand_cluster[j, :3])]
# append back the doppler
hand_cluster = np.append(hand_cluster, doppler_array, 1)
else:
hand_cluster = data
hand_cluster = np.array(hand_cluster)
# apply augmentation to hand cluster #############################
if hand_cluster.size != 0:
# apply augmentations
if 'trans' in augmentation:
for p in np.nditer(hand_cluster[:, :3],
op_flags=['readwrite']):
p[...] = p + random.choice(seeds)
if 'rot' in augmentation:
hand_cluster[:, :3] = rotateX(hand_cluster[:, :3],
720 * random.choice(seeds))
hand_cluster[:, :3] = rotateY(hand_cluster[:, :3],
720 * random.choice(seeds))
hand_cluster[:, :3] = rotateZ(hand_cluster[:, :3],
720 * random.choice(seeds))
if 'scale' in augmentation:
s = 1 + random.choice(seeds)
hand_cluster[:, :3] = scale(hand_cluster[:, :3], x=s, y=s, z=s)
if is_plot:
ax3 = plt.subplot(2, 2, 3, projection='3d')
ax3.set_xlim((-0.3, 0.3))
ax3.set_ylim((-0.3, 0.3))
ax3.set_zlim((-0.3, 0.3))
ax3.set_xlabel('X', fontsize=10)
ax3.set_ylabel('Y', fontsize=10)
ax3.set_zlabel('Z', fontsize=10)
ax3.set_title('Hand Cluster', fontsize=10)
ax3.scatter(hand_cluster[:, 0],
hand_cluster[:, 1],
hand_cluster[:, 2],
'o',
c=hand_cluster[:, 3],
s=28,
marker='o')
# create 3D feature space #############################
frame_3D_volume = snapPointsToVolume(hand_cluster,
volume_shape,
isClipping=('clipping'
in augmentation))
volumes_for_this_interval.append(
np.expand_dims(frame_3D_volume, axis=0))
# Plot the hand cluster #########################################
#################################################################
# Combine the three images
if is_plot:
plt.savefig(
os.path.join(radar_3dscatter_path,
str(timestamp) + '.jpg'))
radar_3dscatter_img = Image.open(
os.path.join(radar_3dscatter_path,
str(timestamp) + '.jpg'))
images = [closest_video_img, radar_3dscatter_img
] # add image here to arrange them horizontally
widths, heights = zip(*(i.size for i in images))
total_width = sum(widths)
max_height = max(heights)
new_im = Image.new('RGB', (total_width, max_height))
x_offset = 0
for im in images:
new_im.paste(im, (x_offset, 0))
x_offset += im.size[0]
timestamp_difference = abs(
float(timestamp) - float(closest_video_timestamp))
draw = ImageDraw.Draw(new_im)
# draw the timestamp difference on the image
(x, y) = (20, 10)
message = "Timestamp Difference, abs(rt-vt): " + str(
timestamp_difference)
draw.text((x, y), message, fill=white_color, font=fnt)
# draw the timestamp
(x, y) = (20, 30)
message = "Timestamp: " + str(timestamp)
draw.text((x, y), message, fill=white_color, font=fnt)
# draw the number of points
(x, y) = (20, 60)
message = "Number of detected points: " + str(xyz.shape[0])
draw.text((x, y), message, fill=white_color, font=fnt)
# draw the number of clusters and number of noise point on the clutter plot
if isCluster:
(x, y) = (20, 80)
message = "Number of clusters: " + str(n_clusters_)
draw.text((x, y), message, fill=white_color, font=fnt)
(x, y) = (20, 100)
message = "Number of outliers: " + str(n_noise_)
draw.text((x, y), message, fill=white_color, font=fnt)
# save the combined image
new_im.save(
os.path.join(
mergedImg_path_intervaled,
str(timestamp) + '_' +
str(timestamp.as_integer_ratio()[0]) + '_' +
str(timestamp.as_integer_ratio()[1]) + '_' +
str(interval_index) + '.jpg'))
plt.close('all')
# process the last interval ##########################################################################
if len(volumes_for_this_interval) <= 100:
num_intervaled_samples = len(volumes_for_this_interval)
if num_intervaled_samples < sample_per_interval / 4:
print('Not Enough Data Points, saving')
else:
# decide the label
if interval_index % inter_arg == 1 or interval_index % inter_arg == 2:
this_label = 0
elif interval_index % inter_arg == 3 or interval_index % inter_arg == 4:
this_label = 1 # for label D
elif interval_index % inter_arg == 5 or interval_index % inter_arg == 6:
this_label = 2 # for label L
elif interval_index % inter_arg == 7 or interval_index % inter_arg == 8:
this_label = 3 # for label M
elif interval_index % inter_arg == 9 or interval_index % inter_arg == 10:
this_label = 4 # for label P
elif interval_index % inter_arg == 11 or interval_index % inter_arg == 12:
this_label = 5 # for label D
elif interval_index % inter_arg == 13 or interval_index % inter_arg == 14:
this_label = 6 # for label L
elif interval_index % inter_arg == 15 or interval_index % inter_arg == 16:
this_label = 7 # for label M
elif interval_index % inter_arg == 17 or interval_index % inter_arg == 18:
this_label = 8 # for label P
elif interval_index % inter_arg == 19 or interval_index % inter_arg == 0:
this_label = 9 # for label P
label_array.append(this_label) # for label A
print('Label for the last interval is ' + str(this_label) +
' Num Samples: ' + str(len(volumes_for_this_interval)))
print('')
# add padding, pre-padded
if len(volumes_for_this_interval) < sample_per_interval:
while len(volumes_for_this_interval) < sample_per_interval:
volumes_for_this_interval.insert(
0, np.expand_dims(np.zeros(volume_shape), axis=0))
elif len(
volumes_for_this_interval
) > sample_per_interval: # we take only the 75 most recent
volumes_for_this_interval = volumes_for_this_interval[-75:]
volumes_for_this_interval = np.asarray(volumes_for_this_interval)
interval_volume_list.append(volumes_for_this_interval)
volumes_for_this_interval = []
# increment the timestamp and interval index
starting_timestamp = starting_timestamp + 5.0
interval_index = interval_index + 1
# start of post processing ##########################################################################
label_array = np.asarray(label_array)
interval_volume_array = np.asarray(interval_volume_list)
# validate the output shapes
assert interval_volume_array.shape == (60, 100, 1) + volume_shape
assert len(label_array) == 60
print('Saving csv and npy to ' + out_path + '...')
if isDataGen:
dataset_path = 'E:/indexPen/dataset'
label_dict_path = 'E:/indexPen/labels/label_dict.p'
# load label dict
if os.path.exists(label_dict_path):
label_dict = pickle.load(open(label_dict_path, 'rb'))
else: # create anew if does not exist
label_dict = {}
# put the label into the dict
for l_index, l in enumerate(label_array):
label_dict[identity_string + str(l_index) + aug_string] = l
# save label dict to disk
pickle.dump(label_dict, open(label_dict_path, 'wb'))
# save the data chunks (intervaled volumns)
for d_index, d in enumerate(interval_volume_array):
print('Saving chunk #' + str(d_index))
np.save(
os.path.join(dataset_path,
identity_string + str(d_index) + aug_string), d)
else:
np.save(os.path.join(out_path, 'label_array'), label_array)
np.save(
os.path.join(
out_path, 'intervaled_3D_volumes_' + str(volume_shape[0]) +
'x' + aug_string), interval_volume_array)
print('Done saving to ' + out_path)
def press(xtext, ytext):
top.destroy()
transformation.scale(img, canvas, float(xtext), float(ytext))
def radar_data_grapher_volumned_track(paths,
isPlot=False,
isCluster=False,
augmentation=(),
seeds=np.random.normal(0, 0.02, 5000),
timesteps=10,
dataset_path='F:/thumouse/dataset/'):
# utility directory to save the pyplots
radarData_path, videoData_path, mergedImg_path, out_path, identity_string = paths
radar_3dscatter_path = 'F:/thumouse/figures/utils/'
radar_data = list(pickle.load(open(radarData_path, 'rb')).items())
radar_data.sort(key=lambda x: x[0]) # sort by timestamp
videoData_list = os.listdir(videoData_path)
videoData_timestamps = list(
map(lambda x: float(x.strip('.jpg')), videoData_list))
style.use('fivethirtyeight')
white_color = 'rgb(255, 255, 255)'
DBSCAN_esp = 0.2
DBSCAN_minSamples = 3
# input data for the classifier that has the shape n*4*100, n being the number of samples
fnt = ImageFont.truetype("arial.ttf", 16)
# removed and recreate the merged image folder
if isPlot:
if os.path.isdir(mergedImg_path):
shutil.rmtree(mergedImg_path)
os.mkdir(mergedImg_path)
volume_shape = (25, 25, 25)
circular_vol_buffer = []
interval_sec = 5
sample_per_sec = 20
sample_per_interval = interval_sec * sample_per_sec
aug_string = ''
if augmentation:
print('Use augmentation: ' + str(augmentation))
for aug in augmentation:
aug_string += '_' + aug
else:
print('No augmentation applied')
for i, radarFrame in enumerate(radar_data):
# retrieve the data
timestamp, fData = radarFrame
if isPlot:
closest_video_timestamp = min(videoData_timestamps,
key=lambda x: abs(x - timestamp))
closest_video_path = os.path.join(
videoData_path,
str(closest_video_timestamp) + '.jpg')
closest_video_img = Image.open(closest_video_path)
# plot the radar scatter
ax1 = plt.subplot(2, 2, 1, projection='3d')
ax1.set_xlim((-0.3, 0.3))
ax1.set_ylim((-0.3, 0.3))
ax1.set_zlim((-0.3, 0.3))
ax1.set_xlabel('X', fontsize=10)
ax1.set_ylabel('Y', fontsize=10)
ax1.set_zlabel('Z', fontsize=10)
ax1.set_title('Detected Points', fontsize=10)
# plot the detected points
ax1.scatter(fData['x'],
fData['y'],
fData['z'],
c=fData['doppler'],
marker='o')
data = np.asarray(
[fData['x'], fData['y'], fData['z'],
fData['doppler']]).transpose()
# Do DBSCAN cluster ###########################################
# map the points to their doppler value, this is for retrieving the doppler value after clustering
if isCluster:
doppler_dict = {}
for point in data:
doppler_dict[tuple(point[:3])] = point[3:]
# get rid of the doppler for clustering TODO should we consider the doppler in clustering?
data = data[:, :3]
db = DBSCAN(eps=DBSCAN_esp,
min_samples=DBSCAN_minSamples).fit(data)
core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
core_samples_mask[db.core_sample_indices_] = True
labels = db.labels_
# Number of clusters in labels, ignoring noise if present.
n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
n_noise_ = list(labels).count(-1)
if isPlot:
ax2 = plt.subplot(2, 2, 2, projection='3d')
ax2.set_xlim((-0.3, 0.3))
ax2.set_ylim((-0.3, 0.3))
ax2.set_zlim((-0.3, 0.3))
ax2.set_xlabel('X', fontsize=10)
ax2.set_ylabel('Y', fontsize=10)
ax2.set_zlabel('Z', fontsize=10)
ax2.set_title('Clustered Points', fontsize=10)
unique_labels = set(labels)
colors = [
plt.cm.Spectral(each)
for each in np.linspace(0, 1, len(unique_labels))
]
clusters = []
for k, col in zip(unique_labels, colors):
if k == -1:
# Black used for noise.
col = [0, 0, 0, 1]
class_member_mask = (labels == k)
xyz = data[class_member_mask & core_samples_mask]
if xyz.any(): # in case there are none objects
clusters.append(
xyz) # append this cluster data to the cluster list
# each cluster is a 3 * n matrix
xyz = data[class_member_mask & ~core_samples_mask]
if isPlot:
ax2.scatter(xyz[:, 0],
xyz[:, 1],
xyz[:, 2],
'o',
c=np.array([col]),
s=12,
marker='X') # plot the noise
# find the center for each cluster
clusters_centers = list(
map(
lambda xyz: np.array([
np.mean(xyz[:, 0]),
np.mean(xyz[:, 1]),
np.mean(xyz[:, 2])
]), clusters))
clusters.sort(
key=lambda xyz: distance.euclidean((0.0, 0.0, 0.0),
np.array([
np.mean(xyz[:, 0]),
np.mean(xyz[:, 1]),
np.mean(xyz[:, 2])
])))
# plot the clusters
for xyz, col in zip(clusters, colors):
if isPlot:
ax2.scatter(xyz[:, 0],
xyz[:, 1],
xyz[:, 2],
'o',
c=np.array([col]),
s=28,
marker='o') # plot the cluster points
#############################
# clear the hand cluster
hand_cluster = []
bbox = (0.2, 0.2, 0.2)
if len(clusters) > 0:
hand_cluster = clusters[0]
point_num = hand_cluster.shape[0]
# if the cluster is outside the 20*20*20 cm bounding box
distance_from_center = distance.euclidean(
(0.0, 0.0, 0.0),
np.array([
np.mean(hand_cluster[:, 0]),
np.mean(hand_cluster[:, 1]),
np.mean(hand_cluster[:, 2])
]))
if distance_from_center > distance.euclidean(
(0.0, 0.0, 0.0), bbox
): # if the core of the cluster is too far away from the center
hand_cluster = np.zeros(
(hand_cluster.shape[0], hand_cluster.shape[1] + 1))
else:
doppler_array = np.zeros((point_num, 1))
for j in range(point_num):
doppler_array[j:, ] = doppler_dict[tuple(
hand_cluster[j, :3])]
# append back the doppler
hand_cluster = np.append(hand_cluster, doppler_array, 1)
else:
hand_cluster = data
hand_cluster = np.array(hand_cluster)
# apply augmentation to hand cluster #############################
if hand_cluster.size != 0:
# apply augmentations
if 'trans' in augmentation:
for p in np.nditer(hand_cluster[:, :3],
op_flags=['readwrite']):
p[...] = p + random.choice(seeds)
if 'rot' in augmentation:
hand_cluster[:, :3] = rotateX(hand_cluster[:, :3],
720 * random.choice(seeds))
hand_cluster[:, :3] = rotateY(hand_cluster[:, :3],
720 * random.choice(seeds))
hand_cluster[:, :3] = rotateZ(hand_cluster[:, :3],
720 * random.choice(seeds))
if 'scale' in augmentation:
s = 1 + random.choice(seeds)
hand_cluster[:, :3] = scale(hand_cluster[:, :3], x=s, y=s, z=s)
if isPlot:
ax3 = plt.subplot(2, 2, 3, projection='3d')
ax3.set_xlim((-0.3, 0.3))
ax3.set_ylim((-0.3, 0.3))
ax3.set_zlim((-0.3, 0.3))
ax3.set_xlabel('X', fontsize=10)
ax3.set_ylabel('Y', fontsize=10)
ax3.set_zlabel('Z', fontsize=10)
ax3.set_title('Hand Cluster', fontsize=10)
ax3.scatter(hand_cluster[:, 0],
hand_cluster[:, 1],
hand_cluster[:, 2],
'o',
c=hand_cluster[:, 3],
s=28,
marker='o')
# create 3D feature space #############################
frame_3D_volume = snapPointsToVolume(hand_cluster,
volume_shape,
isClipping=('clipping'
in augmentation))
print('Processing ' + str(i + 1) + ' of ' + str(len(radar_data)) +
' Circular buffer size: ' + str(len(circular_vol_buffer)))
if timesteps == 1:
this_path = os.path.join(
dataset_path,
str(timestamp.as_integer_ratio()[0]) + '_' +
str(timestamp.as_integer_ratio()[1]) + aug_string)
np.save(this_path, frame_3D_volume)
else:
circular_vol_buffer.append(np.expand_dims(frame_3D_volume, axis=0))
if len(circular_vol_buffer) == timesteps:
# save this sequence
print('saving npy...', end='')
this_path = os.path.join(
dataset_path,
str(timestamp.as_integer_ratio()[0]) + '_' +
str(timestamp.as_integer_ratio()[1]) + aug_string)
if os.path.exists(this_path):
raise Exception(
'File ' + this_path +
' already exists. THIS SHOULD NEVER HAPPEN!')
np.save(this_path, circular_vol_buffer)
print('saved to ' + this_path)
circular_vol_buffer = circular_vol_buffer[1:]
elif len(circular_vol_buffer) > timesteps:
raise Exception(
'Circular Buffer Overflows. THIS SHOULD NEVER HAPPEN!')
# Plot the hand cluster #########################################
#################################################################
# Combine the three images
if isPlot:
plt.savefig(
os.path.join(radar_3dscatter_path,
str(timestamp) + '.jpg'))
radar_3dscatter_img = Image.open(
os.path.join(radar_3dscatter_path,
str(timestamp) + '.jpg'))
images = [closest_video_img, radar_3dscatter_img
] # add image here to arrange them horizontally
widths, heights = zip(*(i.size for i in images))
total_width = sum(widths)
max_height = max(heights)
new_im = Image.new('RGB', (total_width, max_height))
x_offset = 0
for im in images:
new_im.paste(im, (x_offset, 0))
x_offset += im.size[0]
timestamp_difference = abs(
float(timestamp) - float(closest_video_timestamp))
draw = ImageDraw.Draw(new_im)
# draw the timestamp difference on the image
(x, y) = (20, 10)
message = "Timestamp Difference, abs(rt-vt): " + str(
timestamp_difference)
draw.text((x, y), message, fill=white_color, font=fnt)
# draw the timestamp
(x, y) = (20, 30)
message = "Timestamp: " + str(timestamp)
draw.text((x, y), message, fill=white_color, font=fnt)
# draw the number of points
(x, y) = (20, 60)
message = "Number of detected points: " + str(xyz.shape[0])
draw.text((x, y), message, fill=white_color, font=fnt)
# draw the number of clusters and number of noise point on the clutter plot
if isCluster:
(x, y) = (20, 80)
message = "Number of clusters: " + str(n_clusters_)
draw.text((x, y), message, fill=white_color, font=fnt)
(x, y) = (20, 100)
message = "Number of outliers: " + str(n_noise_)
draw.text((x, y), message, fill=white_color, font=fnt)
# save the combined image
new_im.save(
os.path.join(
mergedImg_path,
str(timestamp) + '_' +
str(timestamp.as_integer_ratio()[0]) + '_' +
str(timestamp.as_integer_ratio()[1]) + '.jpg'))
plt.close('all')