def main():
fname = 'data/empire.jpg'
img = Image.open(fname).convert('L')
#img.thumbnail((500, 500))
im = array(img)
pics = [img]
for func in (sobel, gaussian, prewitt):
pics.extend(map(toimg, trans(im, func)))
grid(pics).show()
def main():
fname = 'data/empire.jpg'
img = Image.open(fname).convert('L')
#img.thumbnail((500, 500))
im = array(img)
pics = [img]
for func in (sobel, gaussian, prewitt):
pics.extend(map(toimg, trans(im, func)))
grid(pics).show()
def load_map(self, path):
doc = dom.parse(path)
maptag = doc.firstChild
self.name = os.path.splitext(os.path.basename(path))[0]
self.description = maptag.getAttribute('description')
self.size_str = maptag.getAttribute('size')
self.width = int(self.size_str.split('x')[0])
self.height = int(self.size_str.split('x')[1])
for node in maptag.childNodes:
if node.nodeName == 'Tiles':
self.tiles = str_to_vec_lst(node.firstChild.data)
elif node.nodeName == 'Spawnpoints':
self.spawnpoints = str_to_vec_lst(node.firstChild.data)
elif node.nodeName == 'Portals':
for portal_node in node.childNodes:
if portal_node.nodeName == 'Portal':
# Needs refactoring/improvement
point1 = point2 = p1_dir = p2_dir = None
for p_node in portal_node.childNodes:
if p_node.nodeName == 'p1':
point1 = \
str_to_vec(p_node.firstChild.data)
p1_dir = \
DIRECTIONS[
p_node.getAttribute('dir')]
elif p_node.nodeName == 'p2':
point2 = \
str_to_vec(p_node.firstChild.data)
p2_dir = \
DIRECTIONS[
p_node.getAttribute('dir')]
if point1 is not None and point2 is not None:
self.portals.update(
{point1: (point2, p2_dir),
point2: (point1, p1_dir)})
set_of_tiles = set(grid(self.width, self.height))
set_of_tiles -= set(self.tiles)
accessibility_map = [None] * self.width
accessibility_map = [[None] * self.height for _ in
accessibility_map]
for xpos, ypos in grid(self.width, self.height):
accessibility_map[xpos][ypos] = (xpos, ypos) not in self.tiles
while set_of_tiles:
pos = set_of_tiles.pop()
tiles = flood_fill(accessibility_map,
(self.width, self.height), pos)
portals = []
for portal in self.portals:
if (self.portals[portal][0] in tiles and
portal not in tiles):
portals.append(portal)
self.islands.append(MapAccessibilityNode(tiles, portals))
set_of_tiles -= tiles
directions = np.rad2deg(
np.arctan([(y2 - y1) / ((x2 - x1) + 0.001)
for (x1, y1), (x2, y2) in lines]))
directions = np.where(directions < 0, directions + 180, directions)
hist = np.histogram(directions, range=[0, 180], bins=180)
sort_indexes = np.argsort(hist[0])
hist = hist[0][sort_indexes], hist[1][sort_indexes]
plt.figure()
plt.bar(hist[1], hist[0])
a1, a2 = hist[1][-2:]
print(" - Ângulos mais frequentes: " + str(a1) + ", " + str(a2))
print('Exemplo de grade com frequência e fase especificada')
plt.imshow(utils.grid(pimg.shape, [a1, a2], [0.02, 0.01], [1, 3], 3))
plt.show()
width = 5
img_fit = np.where(pimg == 1, width, -1 / width)
def plot_img_grid(image_grid):
grid_height, grid_width = len(image_grid), len(image_grid[0])
fig, axes = plt.subplots(grid_height,
grid_width,
squeeze=False,
figsize=1.5 *
np.array(plt.rcParams['figure.figsize']))
for i, row in enumerate(axes):
for j, ax in enumerate(row):
def load_map(self, path):
doc = dom.parse(path)
maptag = doc.firstChild
self.name = os.path.splitext(os.path.basename(path))[0]
self.description = maptag.getAttribute('description')
self.size_str = maptag.getAttribute('size')
self.width = int(self.size_str.split('x')[0])
self.height = int(self.size_str.split('x')[1])
for node in maptag.childNodes:
if node.nodeName == 'Tiles':
self.tiles = str_to_vec_lst(node.firstChild.data)
elif node.nodeName == 'Spawnpoints':
self.spawnpoints = str_to_vec_lst(node.firstChild.data)
elif node.nodeName == 'Portals':
for portal_node in node.childNodes:
if portal_node.nodeName == 'Portal':
# Needs refactoring/improvement
point1 = point2 = p1_dir = p2_dir = None
for p_node in portal_node.childNodes:
if p_node.nodeName == 'p1':
point1 = \
str_to_vec(p_node.firstChild.data)
p1_dir = \
DIRECTIONS[
p_node.getAttribute('dir')]
elif p_node.nodeName == 'p2':
point2 = \
str_to_vec(p_node.firstChild.data)
p2_dir = \
DIRECTIONS[
p_node.getAttribute('dir')]
if point1 is not None and point2 is not None:
self.portals.update({
point1: (point2, p2_dir),
point2: (point1, p1_dir)
})
set_of_tiles = set(grid(self.width, self.height))
set_of_tiles -= set(self.tiles)
accessibility_map = [None] * self.width
accessibility_map = [[None] * self.height for _ in accessibility_map]
for xpos, ypos in grid(self.width, self.height):
accessibility_map[xpos][ypos] = (xpos, ypos) not in self.tiles
while set_of_tiles:
pos = set_of_tiles.pop()
tiles = flood_fill(accessibility_map, (self.width, self.height),
pos)
portals = []
for portal in self.portals:
if (self.portals[portal][0] in tiles and portal not in tiles):
portals.append(portal)
self.islands.append(MapAccessibilityNode(tiles, portals))
set_of_tiles -= tiles
#!/usr/bin/env python
from __future__ import print_function
from utils import grid
from utils import partII
if __name__ == '__main__':
power_level = grid(grid_serial_number=1788, size=300)
coordinates, power, size = partII(power_level)
print('Answer: {},{},{}'.format(coordinates[0], coordinates[1], size))
#142,265,7
(0, 0): 0.0,
(1, 0): 0.1,
(2, 0): 0.2,
(3, 0): 0.3,
(4, 0): 0.4,
(5, 0): 0.5,
(6, 0): 0.6,
(7, 0): 0.7,
(8, 0): 0.8,
(9, 0): 0.9,
(10, 0): 1.0,
}
png.write(args.out, config, valResults, 11, 1)
sys.exit(0)
i = 0
for loc in utils.grid(bounds, lat_steps, lng_steps):
yi = i // lng_steps
xi = i % lng_steps
#print('xi = ' + str(xi) + ' yi = ' + str(yi))
y = loc['loc']['lat']
x = loc['loc']['lng']
results[(y, x)] = {}
if (1):
if (funcRecord.get('module')):
results[(y, x)]['val'] = evaluate_variable((y, x), data, config, args.func)
else:
results[(y, x)]['val'] = evaluate((y, x), data, config, funcRecord)
valResults[(xi,yi)] = results[(y, x)]['val']
else:
# for testing with something like "wrmap.py -resolution 5 -eval"
results[(y, x)]['val'] = 'not eval'
def auxGrid(freq1, phase1, freq2, phase2):
return utils.grid(img_fit.shape, [a1, a2], [freq1, freq2],
[phase1, phase2], width)
directions = np.rad2deg(
np.arctan([(y2 - y1) / ((x2 - x1) + 0.001)
for (x1, y1), (x2, y2) in lines]))
directions = np.where(directions < 0, directions + 180, directions)
hist = np.histogram(directions, range=[0, 180], bins=180)
sort_indexes = np.argsort(hist[0])
hist = hist[0][sort_indexes], hist[1][sort_indexes]
plt.figure()
plt.bar(hist[1], hist[0])
a1, a2 = hist[1][-2:]
print(" - Ângulos mais frequentes: " + str(a1) + ", " + str(a2))
print('Exemplo de grade com frequência e fase especificada')
plt.imshow(utils.grid(pimg.shape, [a1, a2], [0.02, 0.01], [1, 3], 3))
plt.show()
width = 5
img_fit = np.where(pimg == 1, width, -1 / width)
def auxGrid(freq1, phase1, freq2, phase2):
return utils.grid(img_fit.shape, [a1, a2], [freq1, freq2],
[phase1, phase2], width)
def denormalize(x):
min_freq = 0.0016521614890856475
high = np.array([0.5, 500, 0.5, 500])
low = np.array([min_freq, 0, min_freq, 0])
def train(train_loader, model, criterion, optimizer, epoch, scheduler):
batch_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
# switch to train mode
model.train()
end = time.time()
prefetcher = data_prefetcher(train_loader)
input, target = prefetcher.next()
i = 0
while input is not None:
i += 1
if args.prof >= 0 and i == args.prof:
print("Profiling begun at iteration {}".format(i))
torch.cuda.cudart().cudaProfilerStart()
if args.prof >= 0:
torch.cuda.nvtx.range_push("Body of iteration {}".format(i))
adjust_learning_rate(scheduler, optimizer, epoch, i, len(train_loader))
if args.grid:
input = grid(input)
# compute output
if args.prof >= 0: torch.cuda.nvtx.range_push("forward")
output = model(input)
if args.prof >= 0: torch.cuda.nvtx.range_pop()
loss = criterion(output, target)
# compute gradient and do SGD step
optimizer.zero_grad()
if args.prof >= 0: torch.cuda.nvtx.range_push("backward")
with amp.scale_loss(loss, optimizer) as scaled_loss:
scaled_loss.backward()
if args.prof >= 0: torch.cuda.nvtx.range_pop()
# for param in model.parameters():
# print(param.data.double().sum().item(), param.grad.data.double().sum().item())
if args.prof >= 0: torch.cuda.nvtx.range_push("optimizer.step()")
optimizer.step()
if args.prof >= 0: torch.cuda.nvtx.range_pop()
if i % args.print_freq == 0:
# Every print_freq iterations, check the loss, accuracy, and speed.
# For best performance, it doesn't make sense to print these metrics every
# iteration, since they incur an allreduce and some host<->device syncs.
# Measure accuracy
prec1, prec5 = accuracy(output.data, target, topk=(1, 5))
# Average loss and accuracy across processes for logging
if args.distributed:
reduced_loss = reduce_tensor(loss.data)
prec1 = reduce_tensor(prec1)
prec5 = reduce_tensor(prec5)
else:
reduced_loss = loss.data
# to_python_float incurs a host<->device sync
losses.update(to_python_float(reduced_loss), input.size(0))
top1.update(to_python_float(prec1), input.size(0))
top5.update(to_python_float(prec5), input.size(0))
torch.cuda.synchronize()
batch_time.update((time.time() - end) / args.print_freq)
end = time.time()
if args.local_rank == 0:
print('Epoch: [{0}][{1}/{2}]\t'
'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Speed {3:.3f} ({4:.3f})\t'
'Loss {loss.val:.10f} ({loss.avg:.4f})\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
'Prec@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
epoch,
i,
len(train_loader),
args.world_size * args.batch_size / batch_time.val,
args.world_size * args.batch_size / batch_time.avg,
batch_time=batch_time,
loss=losses,
top1=top1,
top5=top5))
if args.prof >= 0: torch.cuda.nvtx.range_push("prefetcher.next()")
input, target = prefetcher.next()
if args.prof >= 0: torch.cuda.nvtx.range_pop()
# Pop range "Body of iteration {}".format(i)
if args.prof >= 0: torch.cuda.nvtx.range_pop()
if args.prof >= 0 and i == args.prof + 10:
print("Profiling ended at iteration {}".format(i))
torch.cuda.cudart().cudaProfilerStop()
quit()