def buf() :
current_dset = dataset.dataset('data/basil/front', dtype=float)
comp_dset = dataset.dataset('data/basil/front', dtype=float)
kernel = np.array([[0,0,0],
[0,18,0],
[0,0,0]], dtype=float)
time_kernel = np.array([[-1,-1,-1],
[-1,-1,-1],
[-1,-1,-1]], dtype=float)
for i in comp_dset :
i[...] = cv2.filter2D(i, -1, time_kernel)
it = dataset.buffered_iterator(current_dset, 'movement3', bufsize = 30, outtype=np.uint8, start=1, end_offset=1)
for i in it :
i[...] = cv2.filter2D(i, -1, kernel)
i[...] = comp_dset[it.index-1] + i + comp_dset[it.index+1]
mask = ((i > 100) | (i < -100))
i[...][mask] = 0
mask = (i != 0)
i[...][mask] = 255
mask = ((i[:,:,0] == 0) & (i[:,:,1] == 0) & (i[:,:,2] == 0))
mask = np.invert(mask)
i[...][mask] = 255
def buf():
current_dset = dataset.dataset('data/basil/front', dtype=float)
comp_dset = dataset.dataset('data/basil/front', dtype=float)
kernel = np.array([[0, 0, 0], [0, 18, 0], [0, 0, 0]], dtype=float)
time_kernel = np.array([[-1, -1, -1], [-1, -1, -1], [-1, -1, -1]],
dtype=float)
for i in comp_dset:
i[...] = cv2.filter2D(i, -1, time_kernel)
it = dataset.buffered_iterator(current_dset,
'movement3',
bufsize=30,
outtype=np.uint8,
start=1,
end_offset=1)
for i in it:
i[...] = cv2.filter2D(i, -1, kernel)
i[...] = comp_dset[it.index - 1] + i + comp_dset[it.index + 1]
mask = ((i > 100) | (i < -100))
i[...][mask] = 0
mask = (i != 0)
i[...][mask] = 255
mask = ((i[:, :, 0] == 0) & (i[:, :, 1] == 0) & (i[:, :, 2] == 0))
mask = np.invert(mask)
i[...][mask] = 255
def pofconcept():
dset = dataset.dataset('data/basil/front')
buffer = np.copy(dset[2])
buffer -= dset[1]
buffer2 = np.copy(dset[2])
buffer2 = buffer2.astype(float)
buffer2 -= dset[1]
bmask = ((buffer2 > 0) & (buffer2 < 255))
buffer2[bmask] = 0
bmask = (buffer2 != 0)
buffer2[bmask] = 255
bmask = (buffer2[:, :, 0] == 255) & (buffer2[:, :, 1]
== 255) & (buffer2[:, :, 2] == 255)
bmask = np.invert(bmask)
buffer2[bmask] = 0
cvutil.comp_images(buffer, buffer2.astype(np.uint8))
def poc2():
dset = dataset.dataset('data/basil/front', dtype=float)
#h, s, v = cv2.split(cv2.cvtColor(dset[1], cv2.COLOR_BGR2HSV))
past = np.copy(dset[1])
current = np.copy(dset[2])
future = np.copy(dset[3])
kernel = np.array([[0, 0, 0], [0, 18, 0], [0, 0, 0]], dtype=float)
time_kernel = np.array([[-1, -1, -1], [-1, -1, -1], [-1, -1, -1]],
dtype=float)
past = cv2.filter2D(past, -1, time_kernel)
current = cv2.filter2D(current, -1, kernel)
future = cv2.filter2D(future, -1, time_kernel)
result = past + current + future
mask = ((result > 100) | (result < -100))
result[mask] = 0
mask = (result != 0)
result[mask] = 255
mask = ((result[:, :, 0] == 0) & (result[:, :, 1] == 0) &
(result[:, :, 2] == 0))
mask = np.invert(mask)
result[mask] = 255
dset.set_dtype(np.uint8)
cvutil.comp_images(dset[2], result.astype(np.uint8))
def poc() :
dset = dataset.dataset('data/basil/front')
#h, s, v = cv2.split(cv2.cvtColor(dset[1], cv2.COLOR_BGR2HSV))
result = np.copy(dset[1])
kernel = np.array([[0,-1,0],
[-1,5,-1],
[0,-1,0]], dtype=float)
#kernel *= 1/(kernel.shape[0] * kernel.shape[1])
#for p in (b, g, r) :
# p[...] = ndimage.correlate(p, kernel)
# p[...] = np.clip(p, 0, 255)
#v = ndimage.convolve(v, kernel)
result = cv2.filter2D(result, -1, kernel)
#v = np.clip(v, 0, 255)
#result = cv2.merge((h, s, v))
#result = cv2.cvtColor(result, cv2.COLOR_HSV2BGR)
cvutil.comp_images(dset[1], result)
def sparse_optical_flow() :
dset = dataset.dataset('data/basil/front')
tracking = cv2.goodFeaturesToTrack(cv2.cvtColor(dset[1], cv2.COLOR_BGR2GRAY), maxCorners = 100, qualityLevel = 0.3, minDistance = 7, blockSize = 7)
old = None
for d in range(2,len(dset)) :
very_old = old
old = np.copy(tracking)
tracking, status, error = cv2.calcOpticalFlowPyrLK(cv2.cvtColor(dset[d - 1],cv2.COLOR_BGR2GRAY), cv2.cvtColor(dset[d], cv2.COLOR_BGR2GRAY), old, None, winSize=(15,15), maxLevel= 2)
for i in range(old.shape[0]) :
for j in range(old.shape[1]) :
if (very_old != None) :
cv2.line(dset[d-1], (very_old[i][j][0], very_old[i][j][1]), (old[i][j][0], old[i][j][1]), (0,0,255), thickness=4)
cv2.circle(dset[d-1], (old[i][j][0], old[i][j][1]), 4, (255,0,0), thickness=-1)
dset.write_images('optflowbw')
def apply_test():
dset = dataset.dataset('data/basil/front', dtype=float)
def black(image):
image[:, :, 0] = 0
image[:, :, 1] = 0
image[:, :, 2] = 0
dset.apply(black, 'black')
def apply_test() :
dset = dataset.dataset('data/basil/front', dtype=float)
def black(image) :
image[:,:,0] = 0
image[:,:,1] = 0
image[:,:,2] = 0
dset.apply(black, 'black')
def main() :
current_dset = dataset.dataset('data/basil/front', dtype=float)
comp_dset = dataset.dataset('data/basil/front', dtype=float)
kernel = np.array([[0,0,0],
[0,18,0],
[0,0,0]], dtype=float)
time_kernel = np.array([[-1,-1,-1],
[-1,-1,-1],
[-1,-1,-1]], dtype=float)
for i in comp_dset :
i[...] = cv2.filter2D(i, -1, time_kernel)
for i in current_dset :
i[...] = cv2.filter2D(i, -1, kernel)
for i in range(1, len(current_dset) -1 ) :
current_dset[i] = comp_dset[i-1] + current_dset[i] + comp_dset[i+1]
mask = ((current_dset[i] > 100) | (current_dset[i] < -100))
current_dset[i][mask] = 0
mask = (current_dset[i] != 0)
current_dset[i][mask] = 255
mask = ((current_dset[i][:,:,0] == 0) & (current_dset[i][:,:,1] == 0) & (current_dset[i][:,:,2] == 0))
mask = np.invert(mask)
current_dset[i][mask] = 255
current_dset[i] = current_dset[i].astype(np.uint8)
current_dset.set_dtype(np.uint8)
current_dset.write_images('movement3')
def main() :
for s in sources :
print("starting " + s)
dset = dataset.dataset(os.path.join(source_root, s))
final_dest = os.path.join(dest, dest_prefix + s)
iterator = dataset.buffered_iterator(dset, final_dest, video=True, fps=40)
for f in iterator :
dataset.print_status_bar(iterator.index, iterator.end);
pass
print("\nfinished " + s)
def main():
dset = dataset.dataset('data/basil/front')
for i in dset:
converted = cv2.cvtColor(i, cv2.COLOR_BGR2HSV)
h, s, v = cv2.split(converted)
mask = (h > 45) & (h < 90)
mask = np.invert(mask)
v[mask] = 0
converted = cv2.merge([h, s, v])
i[...] = cv2.cvtColor(converted, cv2.COLOR_HSV2BGR)
dset.write_images('colorf')
def main():
current_dset = dataset.dataset('data/basil/front', dtype=float)
comp_dset = dataset.dataset('data/basil/front', dtype=float)
kernel = np.array([[0, 0, 0], [0, 18, 0], [0, 0, 0]], dtype=float)
time_kernel = np.array([[-1, -1, -1], [-1, -1, -1], [-1, -1, -1]],
dtype=float)
for i in comp_dset:
i[...] = cv2.filter2D(i, -1, time_kernel)
for i in current_dset:
i[...] = cv2.filter2D(i, -1, kernel)
for i in range(1, len(current_dset) - 1):
current_dset[i] = comp_dset[i - 1] + current_dset[i] + comp_dset[i + 1]
mask = ((current_dset[i] > 100) | (current_dset[i] < -100))
current_dset[i][mask] = 0
mask = (current_dset[i] != 0)
current_dset[i][mask] = 255
mask = ((current_dset[i][:, :, 0] == 0) &
(current_dset[i][:, :, 1] == 0) &
(current_dset[i][:, :, 2] == 0))
mask = np.invert(mask)
current_dset[i][mask] = 255
current_dset[i] = current_dset[i].astype(np.uint8)
current_dset.set_dtype(np.uint8)
current_dset.write_images('movement3')
def main() :
dset = dataset.dataset('data/basil/front')
for i in dset :
converted = cv2.cvtColor(i, cv2.COLOR_BGR2HSV)
h, s, v = cv2.split(converted)
mask = (h > 45) & (h < 90)
mask = np.invert(mask)
v[mask] = 0
converted = cv2.merge([h,s,v])
i[...] = cv2.cvtColor(converted, cv2.COLOR_HSV2BGR)
dset.write_images('colorf')
def vein_detect():
dset = dataset.dataset('data/basil/front', dtype=float)
big_med = np.array([
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 0, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
],
dtype=float)
big_sharp = np.array([
[-1, 0, 0, -1, 0, 0, -1],
[0, -1, 0, -1, 0, -1, 0],
[0, 0, -1, -1, -1, 0, 0],
[-1, -1, -1, 25, -1, -1, -1],
[0, 0, -1, -1, -1, 0, 0],
[0, -1, 0, -1, 0, -1, 0],
[-1, 0, 0, -1, 0, 0, -1],
],
dtype=float)
big_med *= (1 / 48)
for image in dset:
sharp_image = cv2.filter2D(image, -1, big_sharp)
blur_image = cv2.filter2D(image, -1, big_med)
sharp_image = np.clip(sharp_image, 0, 255)
blur_image = np.clip(blur_image, 0, 255)
result2 = sharp_image - blur_image
result2[result2 < -40] = 0
result2[result2 > 40] = 0
result2[result2 != 0] = 255
mask = (result2[..., 0] == 0) & (result2[..., 1]
== 0) & (result2[..., 2] == 0)
mask = np.invert(mask)
result2[mask] = 255
image[...] = result2
dset.write_images('edge')
def main() :
dset = dataset.dataset('data/basil/front')
for i in range(len(dset) - 1, 0, -1) :
dset[i] = dset[i].astype(float)
dset[i] -= dset[i - 1]
threshold(dset[i])
dset[i] = dset[i].astype(np.uint8)
cvutil.comp_images(dset[1], dset[2])
dset.write_images('movement')
def main():
dset = dataset.dataset('data/basil/front')
for i in range(len(dset) - 1, 0, -1):
dset[i] = dset[i].astype(float)
dset[i] -= dset[i - 1]
threshold(dset[i])
dset[i] = dset[i].astype(np.uint8)
cvutil.comp_images(dset[1], dset[2])
dset.write_images('movement')
def poc():
dset = dataset.dataset('data/basil/front')
converted = cv2.cvtColor(dset[1], cv2.COLOR_BGR2HSV)
h, s, v = cv2.split(converted)
mask = (h > 45) & (h < 90)
mask = np.invert(mask)
v[mask] = 0
converted = cv2.merge([h, s, v])
cvutil.display_image(cv2.cvtColor(converted, cv2.COLOR_HSV2BGR))
def poc() :
dset = dataset.dataset('data/basil/front')
converted = cv2.cvtColor(dset[1], cv2.COLOR_BGR2HSV)
h, s, v = cv2.split(converted)
mask = (h > 45) & (h < 90)
mask = np.invert(mask)
v[mask] = 0
converted = cv2.merge([h,s,v])
cvutil.display_image(cv2.cvtColor(converted, cv2.COLOR_HSV2BGR))
def vein_detect() :
dset = dataset.dataset('data/basil/front', dtype=float)
big_med = np.array( [ [1, 1, 1, 1 ,1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 0, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1], ], dtype = float)
big_sharp = np.array( [ [-1, 0, 0, -1 ,0, 0, -1],
[0, -1, 0, -1, 0, -1, 0],
[0, 0, -1, -1, -1, 0, 0],
[-1, -1, -1, 25, -1, -1, -1],
[0, 0, -1, -1, -1, 0, 0],
[0, -1, 0, -1, 0, -1, 0],
[-1, 0, 0, -1, 0, 0, -1], ], dtype = float)
big_med *= (1/48)
for image in dset :
sharp_image = cv2.filter2D(image, -1, big_sharp)
blur_image = cv2.filter2D(image, -1, big_med)
sharp_image = np.clip(sharp_image, 0, 255)
blur_image = np.clip(blur_image, 0, 255)
result2 = sharp_image - blur_image
result2[result2 < -40] = 0
result2[result2 > 40] = 0
result2[result2 != 0] = 255
mask = (result2[...,0] == 0) & (result2[...,1] == 0) & (result2[...,2] == 0)
mask = np.invert(mask)
result2[mask] = 255
image[...] = result2
dset.write_images('edge')
def poc_dense() :
dset = dataset.dataset('data/basil/front')
first = cv2.cvtColor(dset[1],cv2.COLOR_BGR2GRAY, dstCn=1)
last = cv2.cvtColor(dset[2], cv2.COLOR_BGR2GRAY, dstCn=1)
canvas = np.zeros_like(dset[1])
canvas[...,1] = 255
flow = cv2.calcOpticalFlowFarneback(first,last ,flow=None, pyr_scale=0.5,levels= 3, winsize=15, iterations=2, poly_n=5, poly_sigma=1.1, flags=0)
mag, ang = cv2.cartToPolar(flow[...,0], flow[...,1])
canvas[...,0] = ang * 180/np.pi/2
canvas[...,2] = cv2.normalize(mag, None, 0, 255,cv2.NORM_MINMAX)
print(canvas)
cvutil.display_image(cv2.cvtColor(canvas, cv2.COLOR_HSV2BGR))
def poc():
image = dataset.dataset('edge')[2]
orig = np.copy(image)
checker = gen_bound_checker(image)
traces = []
cvutil.display_image(image)
for y in range(image.shape[0]):
for x in range(image.shape[1]):
if not (checker(x, y) or trace.in_any_traces(x, y)):
t = trace(x, y, checker)
t.trace_bounds()
t.fill_bounds()
traces.append(t)
dataset.print_status_bar(y * image.shape[1] + x,
image.shape[0] * image.shape[1])
print('\n coloring')
for i, t in enumerate(traces):
if len(t) < 50:
for p in t.points:
image[p.y][p.x][0] = 200
image[p.y][p.x][1] = 200
image[p.y][p.x][2] = 200
else:
randcolor = (random.choice(range(255)), random.choice(range(255)),
random.choice(range(255)))
for p in t.points:
image[p.y][p.x][0] = randcolor[0]
image[p.y][p.x][1] = randcolor[1]
image[p.y][p.x][2] = randcolor[2]
image[t.init_p.y][t.init_p.x][0] = 0
image[t.init_p.y][t.init_p.x][1] = 0
image[t.init_p.y][t.init_p.x][2] = 255
dataset.print_status_bar(i, len(traces))
cvutil.comp_images(orig, image)
def poc() :
dset = dataset.dataset('data/basil/front')
display = np.copy(dset[-1])
tracking = cv2.goodFeaturesToTrack(cv2.cvtColor(dset[1], cv2.COLOR_BGR2GRAY), maxCorners = 100, qualityLevel = 0.3, minDistance = 7, blockSize = 7)
old = tracking
for i in range(2,len(dset)) :
result, status, error = cv2.calcOpticalFlowPyrLK(dset[i - 1], dset[i], old, None, winSize=(15,15), maxLevel= 2)
old = result
for i in range(tracking.shape[0]) :
for j in range(tracking.shape[1]) :
cv2.circle(display, (old[i][j][0], old[i][j][1]), 4, (0,0,255))
cvutil.display_image(display)
def poc2() :
dset = dataset.dataset('data/basil/front', dtype=float)
#h, s, v = cv2.split(cv2.cvtColor(dset[1], cv2.COLOR_BGR2HSV))
past = np.copy(dset[1])
current = np.copy(dset[2])
future = np.copy(dset[3])
kernel = np.array([[0,0,0],
[0,18,0],
[0,0,0]], dtype=float)
time_kernel = np.array([[-1,-1,-1],
[-1,-1,-1],
[-1,-1,-1]], dtype=float)
past = cv2.filter2D(past, -1, time_kernel)
current = cv2.filter2D(current, -1, kernel)
future = cv2.filter2D(future, -1, time_kernel)
result = past + current + future
mask = ((result > 100) | (result < -100))
result[mask] = 0
mask = (result != 0)
result[mask] = 255
mask = ((result[:,:,0] == 0) & (result[:,:,1] == 0) & (result[:,:,2] == 0))
mask = np.invert(mask)
result[mask] = 255
dset.set_dtype(np.uint8)
cvutil.comp_images(dset[2], result.astype(np.uint8))
def dense_optical_flow() :
dset = dataset.dataset('data/basil/front')
prev = cv2.cvtColor(dset[1],cv2.COLOR_BGR2GRAY)
canvas = np.zeros_like(dset[1])
canvas[...,1] = 255
for i in range(2, len(dset)) :
next = cv2.cvtColor(dset[i], cv2.COLOR_BGR2GRAY)
flow = cv2.calcOpticalFlowFarneback(prev, next, flow=None, pyr_scale=0.5,levels= 3, winsize=15, iterations=2, poly_n=5, poly_sigma=1.1, flags=0)
mag, ang = cv2.cartToPolar(flow[...,0], flow[...,1])
canvas[...,0] = ang * 180/np.pi/2
canvas[...,2] = cv2.normalize(mag, None, 0, 255,cv2.NORM_MINMAX)
prev = np.copy(next)
dset[i] = cv2.cvtColor(np.copy(canvas), cv2.COLOR_HSV2BGR)
dset.write_images('denseflow')
def poc() :
image = dataset.dataset('edge')[2]
orig = np.copy(image)
checker = gen_bound_checker(image)
traces = []
cvutil.display_image(image)
for y in range(image.shape[0]) :
for x in range(image.shape[1]) :
if not (checker(x, y) or trace.in_any_traces(x, y)) :
t = trace(x, y, checker)
t.trace_bounds()
t.fill_bounds()
traces.append(t)
dataset.print_status_bar(y * image.shape[1] + x, image.shape[0] * image.shape[1])
print('\n coloring')
for i,t in enumerate(traces) :
if len(t) < 50 :
for p in t.points :
image[p.y][p.x][0] = 200
image[p.y][p.x][1] = 200
image[p.y][p.x][2] = 200
else :
randcolor = (random.choice(range(255)),random.choice(range(255)),random.choice(range(255)))
for p in t.points :
image[p.y][p.x][0] = randcolor[0]
image[p.y][p.x][1] = randcolor[1]
image[p.y][p.x][2] = randcolor[2]
image[t.init_p.y][t.init_p.x][0] = 0
image[t.init_p.y][t.init_p.x][1] = 0
image[t.init_p.y][t.init_p.x][2] = 255
dataset.print_status_bar(i, len(traces))
cvutil.comp_images(orig, image)
def pofconcept() :
dset = dataset.dataset('data/basil/front')
buffer = np.copy(dset[2])
buffer -= dset[1]
buffer2 = np.copy(dset[2])
buffer2 = buffer2.astype(float)
buffer2 -= dset[1]
bmask = ((buffer2 > 0) & (buffer2 < 255))
buffer2[bmask] = 0
bmask = (buffer2 != 0)
buffer2[bmask] = 255
bmask = (buffer2[:,:,0] == 255) & (buffer2[:,:,1] == 255) & (buffer2[:,:,2] == 255)
bmask = np.invert(bmask)
buffer2[bmask] = 0
cvutil.comp_images(buffer, buffer2.astype(np.uint8))
def poc():
dset = dataset.dataset('data/basil/front')
#h, s, v = cv2.split(cv2.cvtColor(dset[1], cv2.COLOR_BGR2HSV))
result = np.copy(dset[1])
kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]], dtype=float)
#kernel *= 1/(kernel.shape[0] * kernel.shape[1])
#for p in (b, g, r) :
# p[...] = ndimage.correlate(p, kernel)
# p[...] = np.clip(p, 0, 255)
#v = ndimage.convolve(v, kernel)
result = cv2.filter2D(result, -1, kernel)
#v = np.clip(v, 0, 255)
#result = cv2.merge((h, s, v))
#result = cv2.cvtColor(result, cv2.COLOR_HSV2BGR)
cvutil.comp_images(dset[1], result)
def poc() :
image = dataset.dataset('data//basil/front')[2]
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(image, 100,200)
cvutil.display_image(cv2.cvtColor(edges, cv2.COLOR_GRAY2BGR))
