def totalDistAndVar(self, groups, dist, varFactor=0):
row_num = len(groups)
combins = list(combinations(list(range(0, row_num)), 2))
#instantiate variable to hold total distance
total_dist = 0
dist_list = []
for i in range(0, len(combins)):
temp_dist_df = groups[groups.iloc[:, 0].isin(combins[i])]
#Get pairwise distance between two groups at a time, using self.pairDist
temp_dist = self.pairDist(temp_dist_df.iloc[0][1:],
temp_dist_df.iloc[1][1:], dist)
dist_list.append(temp_dist)
#add to total_dist
total_dist = total_dist + temp_dist
dist_var = var(dist_list)
total_dist = total_dist + varFactor * dist_var
return dist_var, total_dist
def CalculateVariances(data, classification, averages, count):
variancesLW = []
variancesLD = []
variancesRW = []
variancesRD = []
for scale in data:
if scale.classification == classification:
variancesLW.append(scale.left_weight)
variancesLD.append(scale.left_distance)
variancesRW.append(scale.right_weight)
variancesRD.append(scale.right_distance)
return [
var(variancesLW),
var(variancesLD),
var(variancesRW),
var(variancesRD)
]
def CalculateVariances(data, species, averages, count):
variancesSL = []
variancesSW = []
variancesPL = []
variancesPW = []
for iris in data:
if iris.species == species:
variancesSL.append(iris.sepal_length)
variancesSW.append(iris.sepal_width)
variancesPL.append(iris.petal_length)
variancesPW.append(iris.petal_width)
return [
var(variancesSL),
var(variancesSW),
var(variancesPL),
var(variancesPW)
]
def cdm(x, y, z):
"""
# Multilabel Data
:param x: features
:param y: labels
:param z:
:return: correlation distance matrix
"""
cdm = [] # cdm : final correlation distance matrix
for i in range(len(x[0])):
fea = x[:, i] # fea : features
meanf = sum(fea) / len(fea)
# meanf : mean of features
corr = [] # corr : storing correlation
for j in range(len(y[0])):
lab = y[:, j]
# lab: Labels
print()
meanl = sum(lab) / len(lab)
# meanl : mean length of labels
varx = (var(fea, meanf) * var(lab, meanl)) ** 0.5
# varx : variance of feature
print(varx)
covx = np.cov(fea, lab)[0, 1]
# covx : covariance feature and labels
corr.append((1 - (covx / varx)))
# formula : "1 - (covx / varx)" for finding correlation using covariance and variance
cdm.append(corr) # Adding correlation into cdm
return cdm
import statistics x= statistics.var() y=statistics.alphabets() int=x print(statistics.num(yes))
def test_sd_var2_without_variance(self):
self.assertEqual(Variance("Var2", get_var=False).apply(self.df)["Var2_SD"], [sqrt(var(self.df["Var2"]))])
def test_variance_var2_without_sd(self):
self.assertEqual(Variance("Var2", get_sd=False).apply(self.df)["Var2_Var"], [var(self.df["Var2"])])
def test_projection(idx):
'''
load a point cloud and the corresponding depth map
and the calib matrix
test whether the matrix is working
'''
pc = load_pc(
"/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/velodyne_points/data/"
+ '{:010d}'.format(idx) + ".bin")
vcR, vcT = load_calib(
"/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/calib/calib_velo_to_cam.txt"
)
vc = padRT(vcR, vcT)
proj = load_proj(
"/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/calib/calib_cam_to_cam.txt"
)
depth = load_depth(
"/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/depth/image_02/"
+ '{:010d}'.format(idx) + ".png")
img = load_img(
"/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/image/data/"
+ '{:010d}'.format(idx) + ".png")
'''
visualize_pc(pc)
cv2.imshow('lol',depth/100.0)
cv2.waitKey(0)
neg_pc=list()
pos_pc=list()
zero_pc=list()
neg_count=0
pos_count=0
zero_count=0
for i in range(pc.shape[0]):
if pc[i][0]<0:
neg_count+=1
neg_pc.append(pc[i])
elif pc[i][0]>0:
pos_count+=1
pos_pc.append(pc[i])
else:
zero_count+=1
zero_pc.append(pc[i])
print("neg",neg_count)
print("pos",pos_count)
print("zero",zero_count)
neg_pc=np.stack(neg_pc)
pos_pc=np.stack(pos_pc)
zero_pc=np.stack(zero_pc)
visualize_pc(neg_pc)
visualize_pc(pos_pc)
visualize_pc(zero_pc)
'''
pc_c = list()
pc = euc2hom(pc) # nx4
h, w = depth.shape[0], depth.shape[1]
for i in range(pc.shape[0]):
#print(pc[i])
pc[i] = np.matmul(vc, pc[i])
if pc[i][2] > 0:
#if pc[i][2] != 0:
cp = pc[i]
d = cp[2]
ip = np.matmul(proj, cp)
d = ip[2]
ip = ip / ip[2]
ip[2] = d
#ip[2]=pc[i][2]
pc_c.append(ip)
pc_c = np.stack(pc_c)
print(pc_c.shape)
count = 0
for r in depth:
for e in r:
if e > 0:
count += 1
print(count)
print(np.min(depth))
#pc_corig=pc_c.copy()
#pc_c=hom2euc(pc_c) # nx2
count = 0
diffs = list()
depth2 = np.zeros_like(depth)
for i in range(pc_c.shape[0]):
p = pc_c[i]
#print(p)
x, y = int(round(p[0])), int(round(p[1]))
#print(x,y)
if 0 <= x and 0 <= y and x < w and y < h:
depth2[y, x] = p[2]
if 2 <= x and 2 <= y and x < w - 2 and y < h - 2:
if np.max(depth[y - 2:y + 3, x - 2:x + 3]) == 0:
continue
#dep=np.sum(depth[y-1:y+2,x-1:x+2])/float(np.count_nonzero(depth[y-1:y+2,x-1:x+2]))
d = np.min(np.abs(depth[y - 2:y + 3, x - 2:x + 3] - p[2]))
#if abs(d)>2 and abs(d-p[2])<0.001:
if abs(d) > 5:
count += 1
depth2[y, x] = 512
#print(d,p[2],depth[y-2:y+3,x-2:x+3])
diffs.append(d)
#print("diff:",abs(d))
#print()
'''
'''
print(len(diffs))
print(count)
print(min(diffs))
print(max(diffs))
print(mean(diffs))
print(var(diffs))
cv2.imshow('img', img)
cv2.imshow('lol2', depth2 / 200.0)
cv2.waitKey(0)
for t in tensor:
ret.append(convert_to_colormap(t).unsqueeze(0))
ret = torch.cat(ret)
return ret
if __name__ == '__main__':
#download_kitti()
#process_kitti_raw()
#process_kitti_integ()
#visualize_depth("/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/depth/image_02/0000000136.png")
#load_img("/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/image/data/0000000136.png")
#diffs=load_timediff("/home/chenziwe/robotics/egoDepth/datasets/kitti/2011_09_26_drive_0022_sync/image/timestamps.txt")
'''
'''
dx = list()
dy = list()
dz = list()
for i in range(100):
i + 250
x, y, z = get_diff_means(i, i, velocity=False)
dx.append(x)
dy.append(y)
dz.append(z)
print(mean(dx), mean(dy), mean(dz))
print(var(dx), var(dy), var(dz))
#integ_sequence('2011_09_26_drive_0022')
#x,y,z=get_diff_means(0,400,velocity=True)
#test_projection(5)
#test_reverse(5)
def coeffvar(data_frame):
return (var(data_frame)**(1/2))/mean(data_frame)