def plot_poly_gate(data, gate, ax, chan=None, name=None, **kwargs):
if chan is None:
chan = gate.chan
if 'bgc' in kwargs:
bgc = kwargs['bgc']
del kwargs['bgc']
else:
bgc = 'grey'
if 'bgalpha' in kwargs:
bga = kwargs['bgalpha']
del kwargs['bgalpha']
else:
bga = 1
if 'c' in kwargs:
z = kwargs['c']
del kwargs['c'] # needed incase bgc is used
calc_z = False
else:
calc_z = True
if 'alpha' in kwargs:
alpha = kwargs['alpha']
del kwargs['alpha']
else:
# is there a way to look up what this defaults to incase this is
# changed in .matplotlibrc?
alpha = 1
# add in support for specifing size
ax.scatter(data[:, chan[0]],
data[:, chan[1]],
c=bgc,
s=1,
edgecolor='none',
alpha=bga)
gate.gate(data, chan=chan, name=name)
# has to be set after gating...
if data.shape[0] > 0:
if calc_z:
z = bilinear_interpolate(data[:, chan[0]], data[:, chan[1]])
ax.scatter(data[:, chan[0]],
data[:, chan[1]],
c=z,
s=1,
edgecolor='none',
alpha=alpha,
**kwargs)
ax.fill(gate.vert.T[0],
gate.vert.T[1],
edgecolor='black',
facecolor='none')
def pseudocolor(fcm,
indices,
nrows=1,
ncols=1,
s=1,
edgecolors='none',
savefile=None,
display=True,
**kwargs):
"""Plot a pseudocolor.
indices = list of marker index pairs
nrows = number of rows to plot
ncols = number of cols to plot
nrows * ncols should be >= len(indices)
display = boolean indicating whether to show plot
save = filename to save plot (e.g. 'x.png')
**kwargs are passed on to pylab.scatter
"""
if nrows == 1 and ncols == 1:
ncols = len(indices)
assert (nrows * ncols >= len(indices))
figure = pylab.figure(figsize=(ncols * 4, nrows * 4))
for i, idx in enumerate(indices):
pylab.subplot(nrows, ncols, i + 1)
if (idx[0] != idx[1]):
x = fcm[:, idx[0]]
y = fcm[:, idx[1]]
if 'c' not in kwargs:
z = bilinear_interpolate(x, y)
pylab.scatter(x, y, c=z, s=s, edgecolors=edgecolors, **kwargs)
else:
pylab.scatter(x, y, s=s, edgecolors=edgecolors, **kwargs)
if isinstance(idx[0], str):
pylab.xlabel(idx[0])
else:
pylab.xlabel(fcm.channels[idx[0]])
if isinstance(idx[1], str):
pylab.ylabel(idx[1])
else:
pylab.ylabel(fcm.channels[idx[1]])
pylab.xticks([])
pylab.yticks([])
pylab.axis('equal')
if display:
pylab.show()
if savefile:
pylab.savefig(savefile)
return figure
def pseudocolor(fcm, indices, nrows=1, ncols=1, s=1, edgecolors='none',
savefile=None, display=True,
**kwargs):
"""Plot a pseudocolor.
indices = list of marker index pairs
nrows = number of rows to plot
ncols = number of cols to plot
nrows * ncols should be >= len(indices)
display = boolean indicating whether to show plot
save = filename to save plot (e.g. 'x.png')
**kwargs are passed on to pylab.scatter
"""
if nrows == 1 and ncols == 1:
ncols = len(indices)
assert(nrows * ncols >= len(indices))
figure = pylab.figure(figsize=(ncols * 4, nrows * 4))
for i, idx in enumerate(indices):
pylab.subplot(nrows, ncols, i + 1)
if (idx[0] != idx[1]):
x = fcm[:, idx[0]]
y = fcm[:, idx[1]]
if 'c' not in kwargs:
z = bilinear_interpolate(x, y)
pylab.scatter(x, y, c=z, s=s, edgecolors=edgecolors, **kwargs)
else:
pylab.scatter(x, y, s=s, edgecolors=edgecolors, **kwargs)
if isinstance(idx[0], str):
pylab.xlabel(idx[0])
else:
pylab.xlabel(fcm.channels[idx[0]])
if isinstance(idx[1], str):
pylab.ylabel(idx[1])
else:
pylab.ylabel(fcm.channels[idx[1]])
pylab.xticks([])
pylab.yticks([])
pylab.axis('equal')
if display:
pylab.show()
if savefile:
pylab.savefig(savefile)
return figure
def plot_threshold_gate(data, gate, ax, chan=None, name=None, **kwargs):
if 'bgc' in kwargs:
bgc = kwargs['bgc']
del kwargs['bgc']
else:
bgc = 'grey'
if 'bgalpha' in kwargs:
bga = kwargs['bgalpha']
del kwargs['bgalpha']
else:
bga = 1
if 'c' in kwargs:
z = kwargs['c']
del kwargs['c'] # needed incase bgc is used
calc_z = False
else:
calc_z = True
if 'alpha' in kwargs:
alpha = kwargs['alpha']
del kwargs['alpha']
else:
alpha = 1 # is there a way to look up what this defaults to incase this is changed in .matplotlibrc?
# add in support for specifing size
ax.scatter(data[:,chan[0]],data[:,chan[1]], c=bgc, s=1, edgecolor='none', alpha=bga)
gate.gate(data, name=name)
#has to be set after gating...
if data.shape[0] > 0:
if calc_z:
z = bilinear_interpolate(data[:, chan[0]], data[:, chan[1]])
ax.scatter(data[:,chan[0]],data[:,chan[1]], c=z, s=1, edgecolor='none', alpha=alpha, **kwargs)
if chan[0] == gate.chan:
ax.axvline(gate.vert)
elif chan[1] == gate.chan:
ax.axhline(gate.vert)
def plot_poly_gate(data, gate, ax, chan=None, name=None, **kwargs):
if chan is None:
chan = gate.chan
if 'bgc' in kwargs:
bgc = kwargs['bgc']
del kwargs['bgc']
else:
bgc = 'grey'
if 'bgalpha' in kwargs:
bga = kwargs['bgalpha']
del kwargs['bgalpha']
else:
bga = 1
if 'c' in kwargs:
z = kwargs['c']
del kwargs['c'] # needed incase bgc is used
calc_z = False
else:
calc_z = True
if 'alpha' in kwargs:
alpha = kwargs['alpha']
del kwargs['alpha']
else:
# is there a way to look up what this defaults to incase this is
# changed in .matplotlibrc?
alpha = 1
# add in support for specifing size
ax.scatter(
data[
:,
chan[0]],
data[
:,
chan[1]],
c=bgc,
s=1,
edgecolor='none',
alpha=bga)
gate.gate(data, chan=chan, name=name)
# has to be set after gating...
if data.shape[0] > 0:
if calc_z:
z = bilinear_interpolate(data[:, chan[0]], data[:, chan[1]])
ax.scatter(
data[
:,
chan[0]],
data[
:,
chan[1]],
c=z,
s=1,
edgecolor='none',
alpha=alpha,
**kwargs)
ax.fill(
gate.vert.T[0],
gate.vert.T[1],
edgecolor='black',
facecolor='none')
def main(_):
with tf.Graph().as_default():
#Load image and label
x = tf.placeholder(
shape=[None, FLAGS.resizedheight, FLAGS.resizedwidth, 11],
dtype=tf.float32)
img_list = sorted(glob(FLAGS.dataset_dir + '/*.jpg'))
# # Define the model:
with tf.variable_scope("model") as scope:
with tf.name_scope("depth_prediction"):
#with tf.variable_scope("depth_prediction") as scope:
pred_disp, depth_net_endpoints = depth_net(x,
is_training=False)
# predictions, end_points = resnet_v2.resnet_v2_50(x,
# global_pool=False,
# is_training=False
# )
# multilayer = [end_points['model/resnet_v2_50/block4'],
# end_points['model/resnet_v2_50/block2'],
# end_points['model/resnet_v2_50/block1'],
# end_points['model/depth_prediction/resnet_v2_50/block1/unit_3/bottleneck_v2/conv1'],
# end_points['model/depth_prediction/resnet_v2_50/conv1']]
# pred_disp = upconvolution_net(multilayer,is_training=False)
saver = tf.train.Saver([var for var in tf.model_variables()])
#import pdb;pdb.set_trace()
checkpoint = tf.train.latest_checkpoint(FLAGS.checkpoint_dir)
with tf.Session() as sess:
saver.restore(sess, checkpoint)
#import pdb;pdb.set_trace()
for i in range(len(img_list) - 1):
fh = open(img_list[i], 'r')
I = pil.open(fh)
I = np.array(I)
#I = cv2.resize(I,(FLAGS.resizedwidth,FLAGS.resizedheight),interpolation = cv2.INTER_AREA)
fh = open(img_list[i + 1], 'r')
I1 = pil.open(fh)
I1 = np.array(I1)
# fh = open(img_list[i],'r')
# I = pil.open(fh)
# I = np.array(I)
# I1 = I[:,720:,:]
# I = I[:,:720,:]
#I1 = cv2.resize(I1,(FLAGS.resizedwidth,FLAGS.resizedheight),interpolation = cv2.INTER_AREA)
#I = I.resize((224,224),pil.ANTIALIAS)
#I = I/255.0
#import pdb;pdb.set_trace()
#Optical flow
#flow=np.fromfile(FLAGS.dataset_dir+'/2342_2373.flo.flo', dtype=np.float32).reshape( I1.shape[0],I1.shape[1],2)
flow = util.readFlow(FLAGS.dataset_dir + '/z.flo')
#flow = np.zeros_like(flow)
x_coord = np.repeat(
np.reshape(
np.linspace(0, I1.shape[1] - 1, I1.shape[1]),
[1, I1.shape[1]]), I1.shape[0], 0)
y_coord = np.repeat(
np.reshape(
np.linspace(0, I1.shape[0] - 1, I1.shape[0]),
[I1.shape[0], 1]), I1.shape[1], 1)
x_coord = x_coord + flow[:, :, 0]
y_coord = y_coord + flow[:, :, 1]
#Bilinear interpolation
I_warp, _ = util.bilinear_interpolate(
I1, np.reshape(x_coord, -1), np.reshape(y_coord, -1))
I_warp = I_warp.reshape(I1.shape[0], I1.shape[1], 3)
I_warp = I_warp.astype(np.float32)
I = I.astype(np.float32)
I1 = I1.astype(np.float32)
#I_warp = np.zeros_like(I1)
#I1 = np.zeros_like(I1)
#I = np.zeros_like(I1)
inputdata = np.concatenate([I, I1, flow, I_warp], axis=2)
pred = sess.run(pred_disp,
feed_dict={x: inputdata[None, :, :, :]})
#test = np.asarray(pred[4])
import pdb
pdb.set_trace()
#import pdb;pdb.set_trace()
#z = pred[0][0][0][0,:,:,:]
z = pred[0][0, :, :, 0]
#z=cv2.resize(pred[0][0,:,:,:],(FLAGS.image_width,FLAGS.image_height),interpolation = cv2.INTER_CUBIC)
#z = cv2.bilateralFilter(z,9,75,75)
#z=1.0/z#[0][0,:,:,0]
z.astype(np.float32).tofile(FLAGS.output_dir +
img_list[i].split('/')[-1] +
'_z.bin')
print("The %dth frame is processed" % (i))