def create_b64_image(payload):
def buffered_flo(content):
clean = re.sub("data:application/octet-stream;base64,", '', content)
floBytes = io.BytesIO(base64.b64decode(clean))
return io.BufferedReader(floBytes)
img = fz.convert_from_file(buffered_flo(payload['content']))
uv = fz.convert_from_file(buffered_flo(payload['content']), mode='uv')
rgb = save_image(img, payload['name'])
u = save_image(uv[...,0], payload['name']+'.u', cmap='binary')
v = save_image(uv[...,1], payload['name']+'.v', cmap='binary')
return rgb, u, v
def get_optical_flow(first_img, second_image):
first_tensor = torch.FloatTensor(
numpy.array(first_img)[:, :, ::-1].transpose(2, 0, 1).astype(
numpy.float32) * (1.0 / 255.0))
second_tensor = torch.FloatTensor(
numpy.array(second_image)[:, :, ::-1].transpose(2, 0, 1).astype(
numpy.float32) * (1.0 / 255.0))
tensor_output = estimate(first_tensor, second_tensor)
object_output = open(arguments_strOut, 'wb')
numpy.array([80, 73, 69, 72], numpy.uint8).tofile(object_output)
numpy.array(
[tensor_output.size(2), tensor_output.size(1)],
numpy.int32).tofile(object_output)
numpy.array(tensor_output.detach().numpy().transpose(1, 2, 0),
numpy.float32).tofile(object_output)
object_output.close()
files = glob.glob(arguments_strOut)
img = fz.convert_from_file(files[0])
# print('returning flo: ')
return img
def startOf(self): # A slot takes no params
img_list = utils.listDirectory(self.img_dir)
for ind in range(len(img_list) - 1):
print("Running optical flow on:", img_list[ind], img_list[ind+1])
flo_file = os.path.join(self.flow_dir,"{0}.flo".format(ind))
pwc.setupArguments(model=self.of_model, first_img=img_list[ind],
second_img=img_list[ind+1], save_path=flo_file)
pwc.run()
# Transform from flo to png
flow = fz.convert_from_file(flo_file)
Image.fromarray(flow).save(os.path.join(self.flow_dir,"{0}.png".format(ind)))
self.update.emit(ind)
def show_flow(path: str):
rgb = convert_from_file(path)
filename = path.split('/')[-1]
rgb = cv2.putText(rgb,
filename,
org=(50, 50),
fontFace=cv2.FONT_HERSHEY_SIMPLEX,
fontScale=1,
color=(255, 255, 255))
cv2.imshow("", rgb)
return cv2.waitKey()
def vis_flow(flow, img1, img2, name=""):
import cv2
import flowiz as fz
batch_size = img1.shape[0]
for pair_index in range(batch_size):
cv2.imwrite(name+str(pair_index)+'_img1_test.png', img1[pair_index, :, :, ::-1])
cv2.imwrite(name+str(pair_index)+'_img2_test.png', img2[pair_index, :, :, ::-1])
# save a flow file
output_single = flow[pair_index,:,:,:]
out_file = name+str(pair_index)+'_out.flo'
objOutput = open(out_file, 'wb')
np.array([ 80, 73, 69, 72 ], np.uint8).tofile(objOutput)
np.array([ output_single.shape[2], output_single.shape[1] ], np.int32).tofile(objOutput)
np.array(output_single.cpu().detach().numpy().transpose(1, 2, 0), np.float32).tofile(objOutput)
objOutput.close()
# visualize file
out_img = fz.convert_from_file(out_file)
cv2.imwrite(name+str(pair_index)+'_flow_out.png', out_img[:,:,::-1])
def startOf(self):
"""Start optical flow calculation
"""
img_list = utils.listDirectory(self.img_dir)
of_list = [(img_list[ind], img_list[ind + 1], ind)
for ind in range(len(img_list) - 1)]
for ind in range(len(img_list) - 1):
logging.info("Running optical flow on: {0} {1}".format(
img_list[ind], img_list[ind + 1]))
flo_file = os.path.join(self.of_dir, "{0}.flo".format(ind))
pwc.setupArguments(
model=self.of_model,
first_img=img_list[ind],
second_img=img_list[ind + 1],
save_path=flo_file,
)
pwc.run()
# Transform from flo to png
flow = fz.convert_from_file(flo_file)
Image.fromarray(flow).save(
os.path.join(self.of_dir, "{0}.png".format(ind)))
self.update.emit(ind)
def startBackOf(self): # A slot takes no params
"""Start backward optical flow calculation
"""
img_list = utils.listDirectory(self.img_dir)
for ind in reversed(range(len(img_list) - 1)):
logging.info("Running back optical flow on: {0} {1}".format(
img_list[ind], img_list[ind - 1]))
back_flo_file = os.path.join(self.back_of_dir,
"{0}.flo".format(ind))
pwc.setupArguments(
model=self.of_model,
first_img=img_list[ind],
second_img=img_list[ind - 1],
save_path=back_flo_file,
)
pwc.run()
# Transform from flo to png
flow = fz.convert_from_file(back_flo_file)
Image.fromarray(flow).save(
os.path.join(self.back_of_dir, "{0}.png".format(ind)))
self.update.emit(abs(ind - len(img_list)))
def compute_flow(network, inputs):
global f
global initialize
current_frame = np.array(inputs["input_image"])
if initialize:
frame = current_frame
f.append(current_frame)
initialize = False
else:
f.append(current_frame)
prev_frame = f.pop(0)
next_frame = f[0]
tensorFirst = torch.FloatTensor(
np.array(prev_frame)[:, :, ::-1].transpose(2, 0, 1).astype(
np.float32) * (1.0 / 255.0))
tensorSecond = torch.FloatTensor(
np.array(next_frame)[:, :, ::-1].transpose(2, 0, 1).astype(
np.float32) * (1.0 / 255.0))
tensorOutput = estimate(network, tensorFirst, tensorSecond)
objectOutput = tempfile.NamedTemporaryFile(suffix=".flo")
out = str(Path(objectOutput.name))
np.array([80, 73, 69, 72], np.uint8).tofile(objectOutput)
np.array(
[tensorOutput.size(2), tensorOutput.size(1)],
np.int32).tofile(objectOutput)
np.array(tensorOutput.numpy().transpose(1, 2, 0),
np.float32).tofile(objectOutput)
frame = flowiz.convert_from_file(out)
return {"output_image": frame}
def save_flow_file(fname: str):
img = fz.convert_from_file(fname)
plt.imsave('.'.join(fname.split('.')[:-1] + ['png']), img)
import flowiz as fz
import glob
import matplotlib.pyplot as plt
import cv2
files = glob.glob(
'D:/Study/flownet2-pytorch/out2/inference/run.epoch-0-flow-field/000001.flo'
)
img = fz.convert_from_file(files[0])
# cv2.imshow("",img)
# cv2.waitKey(0)
plt.imshow(img)
plt.show()
def plot_optical_vectors(url, cut_co):
vis = fz.convert_from_file(url)
plt.imshow(vis[:, cut_co[0]:cut_co[0] + cut_co[1]])