def disparity_minus_mean():
down_factor = 2
# index = 1
test=False
n_disp = 128
differences = []
border = 10
for index in range(2):
print(index)
source = KittiMultiViewSource(index, test=test, n_frames=0)
average_disparity = source.get_average_disparity()
# frame = load_pair(index, test, frame=i, multiview=True)
frame = source.frame_ten
gt_frame = calc_ground_truth(frame, n_disp, down_factor=down_factor, iters=50)
d = gt_frame.astype('int') - average_disparity.astype('int')
d = d[border:-border,border:-border]
differences.append(d.flatten())
# plt.subplot(211)
# plt.imshow(gt_frame)
# plt.colorbar()
# plt.subplot(212)
# plt.imshow(average_disparity)
# plt.colorbar()
# plt.show()
foo = np.array(differences)
print()
print(np.array(differences).shape)
def disparity_minus_mean():
down_factor = 2
# index = 1
test = False
n_disp = 128
differences = []
border = 10
for index in range(2):
print(index)
source = KittiMultiViewSource(index, test=test, n_frames=0)
average_disparity = source.get_average_disparity()
# frame = load_pair(index, test, frame=i, multiview=True)
frame = source.frame_ten
gt_frame = calc_ground_truth(frame,
n_disp,
down_factor=down_factor,
iters=50)
d = gt_frame.astype('int') - average_disparity.astype('int')
d = d[border:-border, border:-border]
differences.append(d.flatten())
# plt.subplot(211)
# plt.imshow(gt_frame)
# plt.colorbar()
# plt.subplot(212)
# plt.imshow(average_disparity)
# plt.colorbar()
# plt.show()
foo = np.array(differences)
print()
print(np.array(differences).shape)
def foveation_sequence():
frame_down_factor = 1
mem_down_factor = 2 # relative to the frame down factor
coarse_down_factor = 2 # for the coarse comparison
fs = 80
fovea_shape = (fs, fs)
full_values = 128
values = full_values / 2**frame_down_factor
index = 15
n_frames = 10
source = KittiMultiViewSource(index, test=False, n_frames=n_frames)
full_shape = source.frame_ten[0].shape
frame_ten = [downsample(source.frame_ten[0], frame_down_factor),
downsample(source.frame_ten[1], frame_down_factor)]
frame_shape = frame_ten[0].shape
average_disp = source.get_average_disparity()
average_disp = cv2.pyrUp(average_disp)[:frame_shape[0],:frame_shape[1]-values]
filter = Filter(average_disp, frame_down_factor, mem_down_factor,
fovea_shape, frame_shape, values, verbose=False, memory_length=0)
plt.figure()
import matplotlib.cm as cm
for i in range(0, 10, 2):
frame = [downsample(source.frame_sequence[i][0], frame_down_factor),
downsample(source.frame_sequence[i][1], frame_down_factor)]
filter_disp, fovea_corner = filter.process_frame(None, frame)
edge = 5
plt.subplot(5,1,i/2+1)
# plt.subplot(5,2,i+1)
plt.imshow(trim(frame[0], values, edge), cmap = cm.Greys_r)
# remove_axes()
# plt.subplot(5,2,i+2)
# plt.imshow(trim(filter_disp, values, edge), vmin=0, vmax=full_values)
fovea_corner = fovea_corner[0]
# plot_edges(fovea_ij, (fs, fs))
fi, fj = fovea_corner
fm = fs
fn = fs
plt.plot([fj, fj+fn, fj+fn, fj, fj], [fi, fi, fi+fm, fi+fm, fi], 'white')
# plt.scatter(fovea_corner[1]-values+fs/2, fovea_corner[0]-edge+fs/2, s=100, c='green', marker='+', linewidths=2)
# plt.scatter(fovea_corner[1]-values, fovea_corner[0]-edge, s=9, c='green', marker='+', linewidths=3)
# plt.scatter(fovea_corner[1]-values+fs, fovea_corner[0]-edge+fs, s=9, c='green', marker='+', linewidths=3)
# plt.scatter(fovea_corner[1]-values, fovea_corner[0]-edge+fs, s=9, c='green', marker='+', linewidths=3)
# plt.scatter(fovea_corner[1]-values+fs, fovea_corner[0]-edge, s=9, c='green', marker='+', linewidths=3)
remove_axes()
plt.tight_layout(-1)
plt.show()
def objective(args):
# args['ksize'] = int(args['ksize'])
costs = []
# for index in range(3):
# for index in range(10):
for index in range(194):
if index in [31, 82, 114]: # missing frames
continue
source = KittiMultiViewSource(index, test=False, n_frames=0)
full_shape = source.frame_ten[0].shape
frame_ten = [downsample(source.frame_ten[0], frame_down_factor),
downsample(source.frame_ten[1], frame_down_factor)]
frame_shape = frame_ten[0].shape
true_points = source.get_ground_truth_points(occluded=False)
# determine fovea shape
fovea_pixels = np.ceil(fovea_fraction * frame_shape[0] * (frame_shape[1] - values))
if fovea_pixels > 0:
fovea_height = np.minimum(frame_shape[0], np.ceil(fovea_pixels ** .5))
fovea_width = np.minimum(frame_shape[1], np.ceil(fovea_pixels / fovea_height))
fovea_shape = fovea_height, fovea_width
else:
fovea_shape = (0, 0)
# average disparity
try:
average_disp = source.get_average_disparity()
except IOError: # likely does not have a full 20 frames
print("Skipping index %d (lacks frames)" % index)
continue
average_disp = upsample_average_disp(
average_disp, frame_down_factor, frame_shape, values)
# run the filter
foveal = Foveal(average_disp, frame_down_factor, mem_down_factor,
fovea_shape, frame_shape, values,
iters=iters, fovea_levels=fovea_levels, **args)
disp, _ = foveal.process_frame(frame_ten)
# cost = cost_on_points(disp[:,values:], true_points, full_shape=full_shape)
cost = cost_on_points(disp[:,values:], true_points, average_disp,
full_shape=full_shape, clip=error_clip)
costs.append(cost)
error = np.mean(costs)
errors[arg_key(args)] = error
args_s = "{%s}" % ', '.join("%r: %s" % (k, args[k]) for k in sorted(space))
print("%s: %s" % (error, args_s))
return error
#plt.show(block=True)
area = 93 * 275
fovea_area = 0.4 * area
height = np.floor(np.sqrt(fovea_area))
width = np.floor(fovea_area / height)
fovea_shape = (height, width)
print(np.floor(np.sqrt(area)))
n_frames = 194
best_nums = []
for index in range(n_frames):
print(index)
source = KittiMultiViewSource(index)
average_disparity = source.get_average_disparity()
uc = UnusuallyClose(average_disparity)
gt = calc_ground_truth(source.frame_ten, 128, down_factor=2, iters=3)
cost = uc.get_importance(gt)
foveas_ij, fovea_part_shape = _choose_foveas(cost, fovea_shape, values, 30)
best_nums.append(len(foveas_ij))
print('best nums:')
print(best_nums)
plt.figure()
plt.hist(best_nums, np.add(range(31), 0.5))
plt.xlabel('Number of foveas', fontsize=18)
plt.ylabel('Frequency of best cost', fontsize=18)
plt.gca().tick_params(labelsize='18')
plt.xlim((0, 31))
area = 93*275
fovea_area = 0.4 * area
height = np.floor(np.sqrt(fovea_area))
width = np.floor(fovea_area / height)
fovea_shape = (height, width)
print(np.floor(np.sqrt(area)))
n_frames = 194
best_nums = []
for index in range(n_frames):
print(index)
source = KittiMultiViewSource(index)
average_disparity = source.get_average_disparity()
uc = UnusuallyClose(average_disparity)
gt = calc_ground_truth(source.frame_ten, 128, down_factor=2, iters=3)
cost = uc.get_importance(gt)
foveas_ij, fovea_part_shape = _choose_foveas(cost, fovea_shape, values, 30)
best_nums.append(len(foveas_ij))
print('best nums:')
print(best_nums)
plt.figure()
plt.hist(best_nums, np.add(range(31), 0.5))
plt.xlabel('Number of foveas', fontsize=18)
plt.ylabel('Frequency of best cost', fontsize=18)
plt.gca().tick_params(labelsize='18')
plt.xlim((0, 31))
def foveation_sequence():
frame_down_factor = 1
mem_down_factor = 2 # relative to the frame down factor
coarse_down_factor = 2 # for the coarse comparison
fs = 80
fovea_shape = (fs, fs)
full_values = 128
values = full_values / 2**frame_down_factor
index = 15
n_frames = 10
source = KittiMultiViewSource(index, test=False, n_frames=n_frames)
full_shape = source.frame_ten[0].shape
frame_ten = [
downsample(source.frame_ten[0], frame_down_factor),
downsample(source.frame_ten[1], frame_down_factor)
]
frame_shape = frame_ten[0].shape
average_disp = source.get_average_disparity()
average_disp = cv2.pyrUp(average_disp)[:frame_shape[0], :frame_shape[1] -
values]
filter = Filter(average_disp,
frame_down_factor,
mem_down_factor,
fovea_shape,
frame_shape,
values,
verbose=False,
memory_length=0)
plt.figure()
import matplotlib.cm as cm
for i in range(0, 10, 2):
frame = [
downsample(source.frame_sequence[i][0], frame_down_factor),
downsample(source.frame_sequence[i][1], frame_down_factor)
]
filter_disp, fovea_corner = filter.process_frame(None, frame)
edge = 5
plt.subplot(5, 1, i / 2 + 1)
# plt.subplot(5,2,i+1)
plt.imshow(trim(frame[0], values, edge), cmap=cm.Greys_r)
# remove_axes()
# plt.subplot(5,2,i+2)
# plt.imshow(trim(filter_disp, values, edge), vmin=0, vmax=full_values)
fovea_corner = fovea_corner[0]
# plot_edges(fovea_ij, (fs, fs))
fi, fj = fovea_corner
fm = fs
fn = fs
plt.plot([fj, fj + fn, fj + fn, fj, fj],
[fi, fi, fi + fm, fi + fm, fi], 'white')
# plt.scatter(fovea_corner[1]-values+fs/2, fovea_corner[0]-edge+fs/2, s=100, c='green', marker='+', linewidths=2)
# plt.scatter(fovea_corner[1]-values, fovea_corner[0]-edge, s=9, c='green', marker='+', linewidths=3)
# plt.scatter(fovea_corner[1]-values+fs, fovea_corner[0]-edge+fs, s=9, c='green', marker='+', linewidths=3)
# plt.scatter(fovea_corner[1]-values, fovea_corner[0]-edge+fs, s=9, c='green', marker='+', linewidths=3)
# plt.scatter(fovea_corner[1]-values+fs, fovea_corner[0]-edge, s=9, c='green', marker='+', linewidths=3)
remove_axes()
plt.tight_layout(-1)
plt.show()