def main():
desc = argv.next() or "B2"
n = int(desc[-1:])
if desc[0] == "A":
gen = build_An(n)
elif desc[0] == "B":
gen = build_Bn(n)
elif desc[0] == "D":
gen = build_Dn(n)
else:
print("no group desc found matching %s"%desc)
return
I = Op.identity(n)
zero = Op(n)
G = mulclose(gen)
print(len(G))
projs = []
for g in G:
if g==I:
continue
if g*g==I:
projs.append(posproj(g, I))
assert(len(set(projs))==len(projs))
#for H in search(projs):
# print("found")
pairs = []
for p in projs:
for q in projs:
if p*q != q*p:
pairs.append((p, q))
print("pairs:", len(pairs))
shuffle(pairs)
for (p1, q1) in pairs:
for (p2, q2) in pairs:
A = tensor(p1, p2)
B = tensor(q1, q2)
if A*B == B*A:
write(".")
print()
def load_image_rect_flatten_labels(img_path, start_point, end_point, padding,
color = (0, 255, 0), color_pad = (255, 0, 0), thickness = 2):
#Import image
orig_image = cv2.imread(img_path)
#Show the image with matplotlib
start_point_pad = (start_point[0]-padding, start_point[1]-padding)
end_point_pad = (end_point[0]+padding, end_point[1]+padding)
image = orig_image.copy()
image = cv2.rectangle(image, start_point, end_point, color, thickness)
image = cv2.rectangle(image, start_point_pad, end_point_pad, color_pad, thickness)
# label image
image_labels = np.zeros((orig_image.shape[0], orig_image.shape[1]))
image_labels[start_point[1]:end_point[1], start_point[0]:end_point[0]] = 1.0
# crop
img = orig_image[start_point_pad[1]:end_point_pad[1], start_point_pad[0]:end_point_pad[0]]
labels = image_labels[start_point_pad[1]:end_point_pad[1], start_point_pad[0]:end_point_pad[0]]
img = np.moveaxis(img, -1, 0)
img = torch.tensor(img)
img = img.float()
img = img.unsqueeze(0)
labels = np.tensor(labels)
flatten_labels = labels.view(-1)
foreground = torch.where(flatten_labels == 1)[0]
foreground = foreground.tolist()
background = torch.where(flatten_labels == 0)[0]
background = background.tolist()
# Add to return image to see image with rectangles
return img, flatten_labels, foreground, background, image
def calc(model, data_loader):
model.eval()
dataset = data_loader.dataset
prog_bar = mmcv.ProgressBar(len(dataset))
seg_eval = SegEvaluator(class_names=dataset.SEG_CLASSES)
det_eval = SegEvaluator(class_names=dataset.SEG_CLASSES)
print('\nStart Test Loop')
# print('batch_size:', data_loader.batch_size) # 1
for idx, data in enumerate(data_loader):
# print(type(data['img'][0])) # DataContainter
with torch.no_grad():
seg_res, box_res = model(return_loss=False, rescale=True, **data)
# len(box_res) == batch_size == 1
# box_res: [dict('pts_bbox'=dict(boxes_3d=bboxes, scores_3d=scores, labels_3d=labels))]
# handle seg
seg_label = data['seg_label'][0].data[0] # list of tensor
seg_pts_indices = data['seg_pts_indices'][0].data[0] # list of tensor
seg_points = data['seg_points'][0].data[0]
seg_pred = seg_res.argmax(1).cpu().numpy()
pred_list = []
gt_list = []
left_idx = 0
for i in range(len(seg_label)):
# num_points = len(seg_pts_indices[i])
assert len(seg_label[i]) == len(seg_pts_indices[i])
num_points = len(seg_label[i])
right_idx = left_idx + num_points
pred_list.append(seg_pred[left_idx:right_idx])
gt_list.append(seg_label[i].numpy())
left_idx = right_idx
seg_eval.batch_update(pred_list, gt_list)
# handle det
dic = box_res[0]['pts_bbox']
tensor_boxes = dic['boxes_3d'].tensor[:, :7].cuda()
labels = dic['labels_3d']
num_seg_pts = len(seg_points[0])
num_pred_boxes = len(tensor_boxes)
fake_labels = torch.tensor([4] * num_seg_pts)
box_idx = points_in_boxes_gpu(seg_points[0].cuda().unsqueeze(0),
tensor_boxes.unsqueeze(0)).squeeze(0)
for i in range(num_pred_boxes):
mask = box_idx == i # select points in i_th box
fake_labels[mask] = labels[i]
det_eval.update(fake_labels.numpy(), seg_label[0])
# progress bar
batch_size = len(box_res)
for _ in range(batch_size):
prog_bar.update()
print(seg_eval.print_table())
print('overall_acc:', seg_eval.overall_acc)
print('overall_iou:', seg_eval.overall_iou)
print(det_eval.print_table())
print('overall_acc:', det_eval.overall_acc)
print('overall_iou:', det_eval.overall_iou)
def test():
for n in [2, 3, 4]:
ops = build_An(n)
assert len(mulclose(ops))==factorial(n)
ops = build_Bn(n)
assert len(mulclose(ops))==2**n*factorial(n)
ops = build_Dn(n)
assert len(mulclose(ops))==2**(n-1)*factorial(n)
from numpy import kron as tensor
def allclose(A, B):
return numpy.abs(A - B).sum()==0
for A in ops:
for B in ops:
assert (A==B) == (hash(A)==hash(B))
assert (A!=B) != (A==B)
lhs = (A*B).todense()
rhs = ((numpy.dot(A.todense(), B.todense())))
assert allclose(lhs, rhs)
lhs = (A.tensor(B)).todense()
rhs = ((tensor(A.todense(), B.todense())))
assert allclose(lhs, rhs)
for A in ops:
assert allclose(A.transpose().todense(), A.todense().transpose()), str(A)
ops = mulclose(build_Bn(2))
tops = []
for A in ops:
for B in ops:
C = Tensor([A, B])
tops.append(C)
#print(len(tops))
for A in tops:
assert A.get_canonical() == A
for A in tops:
for B in tops:
assert (A==B) == allclose(A.todense(), B.todense())
assert (A==B) == (A.get_canonical()==B.get_canonical())
assert (A==B) == (hash(A)==hash(B))
print("OK")
def test():
from numpy import kron as tensor
perms = [[0, 1, 2], [0, 2, 1], [1, 0, 2], [1, 2, 0], [2, 1, 0], [2, 0, 1]]
signss = [(a, b, c) for a in [-1, 1] for b in [-1, 1] for c in [-1, 1]]
ops = [Op(perm, signs) for perm in perms for signs in signss]
for A in ops:
for B in ops:
assert (A == B) == (hash(A) == hash(B))
assert (A != B) != (A == B)
lhs = (A * B).todense()
rhs = ((numpy.dot(A.todense(), B.todense())))
assert numpy.allclose(lhs, rhs)
lhs = (A.tensor(B)).todense()
rhs = ((tensor(A.todense(), B.todense())))
assert numpy.allclose(lhs, rhs)
for A in ops:
assert numpy.allclose(A.transpose().todense(),
A.todense().transpose()), str(A)
ops = mulclose(build_Bn(2))
tops = []
for A in ops:
for B in ops:
C = Tensor([A, B])
tops.append(C)
#print(len(tops))
for A in tops:
assert A.get_canonical() == A
for A in tops:
for B in tops:
assert (A == B) == numpy.allclose(A.todense(), B.todense())
assert (A == B) == (A.get_canonical() == B.get_canonical())
assert (A == B) == (hash(A) == hash(B))
print("OK")
def multi_toric(Xs, Zs):
n = 8
N = Xs[0].n
I = Op.identity(N)
for X in Xs:
for Z in Zs:
assert X * X == I
assert Z * Z == I
assert Z * X == -X * Z
print("toric: dim=%d" % (N**n))
II = Op.identity(N**n)
plaqs = [(0, 1, 2, 5), (0, 2, 3, 7), (4, 5, 6, 1), (6, 7, 3, 4)]
stars = [(0, 1, 3, 4), (1, 2, 3, 6), (0, 4, 5, 7), (2, 5, 6, 7)]
for a in plaqs:
for b in stars:
assert len(set(a).intersection(set(b))) % 2 == 0
ops = []
for idxs in stars:
for X in Xs:
op = [I] * n
for idx in idxs:
op[idx] = X
op = tensor(*op)
assert op * op == II
ops.append(op)
for idxs in plaqs:
for Z in Zs:
op = [I] * n
for idx in idxs:
op[idx] = Z
op = tensor(*op)
assert op * op == II
ops.append(op)
for A in ops:
for B in ops:
assert A * B == B * A
v = numpy.zeros(N**n)
v[0] = 1.
for A in ops:
v = v + A(v)
assert (abs(norm(v)) > 0.1)
v /= norm(v)
#print(v)
spaces = []
for A in ops:
S = A.get_stab()
#print("get_stab:", len(S))
spaces.append(S)
u = A(v)
r = numpy.dot(u, v)
err = norm(u - r * v)
assert abs(r - 1.) < EPSILON
assert err < EPSILON
#print(r, err, end=' ')
#print()
S = spaces[0]
for S1 in spaces[1:]:
print("codespace:", len(S))
S = S.intersect(S1)
print("codespace:", len(S))
import sys
import GP_function as gp
import GP_function_torch as gpt
import nuts as nuts
#%% read data
data_s = pd.read_csv("Accord_Simulation_Results_LL_Airbag_300_Large_Range.csv")
data_f = pd.read_csv("Field_Test_Data_Full_Size_Sedan.csv")
data_l = pd.read_csv("injury_function_data_chest.csv")
data_e = pd.read_csv("Experimental Data - All Sedans.csv")
n_var = 3
n_para = 7
n_res = 1
parameter_default = tch.tensor([1, 0.45, 1, 1, 0.55, 1, 1])
val_nugget = 10 ^ (-4)
n_s = len(data_s)
x_s = data_s.iloc[:, 1:(n_var + n_para + 1)].to_numpy()
y_s = tch.tensor(data_s.iloc[:, n_var + n_para + 1])
n_f = len(data_f)
x_f = data_f.iloc[:, 3:6].to_numpy()
a_f = data_f['Age'].to_numpy()
z_f = (data_f.iloc[:, 7] >= 3).to_numpy()
n_l = len(data_l)
a_l = data_l['Age'].to_numpy()
z_l = data_l['Injury Indicator'].to_numpy()
y_l = data_l['ChestD'].to_numpy()
def main():
X, Z = build(2)
I = dot(X, X)
#print(I)
#print(X)
#print(Z)
II = tensor(I, I)
P2 = [tensor(X, I), tensor(Z, I), tensor(I, X), tensor(I, Z)]
found = set()
for g in mulclose(P2):
s = shortstr(g)
found.add(s)
assert len(found) == 32
assert shortstr(II) in found
if 0:
G = mulclose(P2)
for g in G:
A = numpy.array(g, dtype=numpy.float)
vals = numpy.linalg.eigvals(A)
print(vals) # all +1, -1
d = argv.get("d", 4)
gen = build(d)
n = (2**d) * factorial(d)
assert len(mulclose(gen)) == n # slow...
G = mulclose(gen, n)
print("orders:")
for g in G:
print(order(g), end=' ')
print()
if 0:
for g in G:
A = numpy.array(g, dtype=numpy.float)
vals = numpy.linalg.eigvals(A)
print(vals) # all kinds of stuff..
if d == 4:
assert len([g for g in G if shortstr(g) in found]) == 32
pairs = []
c_comm = 0
c_anti = 0
total = 0
for i in range(len(G)):
for j in range(len(G)):
g = G[i]
h = G[j]
gh = dot(g, h)
hg = dot(h, g)
total += 1
if eq(gh, -hg):
c_anti += 1
elif eq(gh, hg):
c_comm += 1
if shortstr(g) in found or shortstr(h) in found:
continue
if eq(gh, -hg) and i < j:
#assert order(g)==2
print(order(g), end=' ')
#print(g)
#print(h)
#print(eq(g, g.transpose()), end=' ')
#print(eq(h, h.transpose()))
if eq(g, g.transpose()) and eq(h, h.transpose()) and i < j:
pairs.append((g, h))
#print(".", end='')
#print()
#print(c_comm, c_anti, total)
print("pairs:", len(pairs))
ops = {}
for (g, h) in pairs:
ops[shortstr(g)] = g
ops[shortstr(h)] = h
ops = list(ops.values())
print(len(ops))
for g in ops:
for h in ops:
a = eq(dot(g, h), dot(h, g))
b = eq(dot(g, h), -dot(h, g))
if a == b == False:
s = ' '
elif a:
s = '+ '
elif b:
s = '- '
print(s, end=' ')
print()
return
A, B, C, D = gen
pairs = [
s.split() for s in ("ABDA CBDC", "BACB DCDA", "BACB DCDC",
"BACDBCBADCBC DCABDCDABACD")
]
for l, r in pairs:
left = [eval(s) for s in l]
right = [eval(s) for s in r]
left = dotx(*left)
right = dotx(*right)
if 1:
print(shortstr(left))
print(shortstr(left) in found)
print()
print(shortstr(right))
print(shortstr(right) in found)
print()
print(shortstr(dotx(left, right)))
print()
print()
assert eq((dotx(right, left)), -(dotx(left, right)))
using some bitwise arithmetic voodoo
"""
return [x >> i & 1 for i in range(10)]
def fizz_buzz_encode(x: int) -> List[int]:
if x % 15 == 0:
return [0, 0, 0, 1]
elif x % 5 == 0:
return [0, 0, 1, 0]
elif x % 3 == 0:
return [0, 1, 0, 0]
else:
return [1, 0, 0, 0]
x_train = tensor([binary_encode(i) for i in range(101, 1024)])
y_train = tensor([fizz_buzz_encode(i) for i in range(101, 1024)])
HIDDEN_SIZE = 50
BATCH_SIZE = 32
LEARNING_RATE = 0.001
NUM_EPOCHS = 10_000
net = NeuralNet([
Linear(input_size=10, output_size=HIDDEN_SIZE),
Tanh(),
Linear(input_size=HIDDEN_SIZE, output_size=4)
])
loss = SSE()
optimizer = SGD(lr=LEARNING_RATE)
tensor_boxes = tensor_boxes[:, :7]
box_idx = points_in_boxes_gpu(pts1.unsqueeze(0), tensor_boxes.unsqueeze(0)).squeeze(0)
# boxes = LiDARInstance3DBoxes(gt_bboxes_3d.tensor[:, :7])
# box_idx = boxes.points_in_boxes(seg_points[:, :3].cuda())
# 1.
# fake_labels = torch.tensor([num_classes-1] * len(seg_labels))
# for i in range(len(box_idx)):
# if box_idx[i] != -1:
# fake_labels[i] = gt_labels_3d[box_idx[i]]
# 2.
# fake_labels = torch.tensor(list(map(lambda x: gt_labels_3d[x] if x >= 0 else num_classes-1, box_idx)))
# 3.
fake_labels = torch.tensor([num_classes-1] * len(seg_labels))
mask = box_idx != -1
fake_labels[mask] = torch.tensor(list(map(lambda x: gt_labels_3d[x], box_idx[mask])), dtype=torch.long)
end = time.time()
t1 = end - start
evaluator = SegEvaluator(class_names=dataset.SEG_CLASSES)
evaluator.update(fake_labels.numpy(), seg_labels.numpy())
print(evaluator.print_table())
print('overall_acc:', evaluator.overall_acc)
print('overall_iou:', evaluator.overall_iou)
# nuscenes points_in_box
start = time.time()
allcorners = gt_bboxes_3d.corners # (N, 8, 3)
fake_labels = torch.tensor([num_classes-1] * len(seg_labels))
