def overlay_cells(src,
center_intensity_radius,
dilation_radius=5,
load_range=False):
'''
src = fullsizedata folder
center_intensity_radius = path to center_intensity_radius.p file
src = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20161214_db_bl6_crii_l_53hr/full_sizedatafld/20161214_db_bl6_crii_l_53hr_647_010na_z7d5um_75msec_5POVLP_ch00'
center_intensity_radius = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20161214_db_bl6_crii_l_53hr/cells/20161214_db_bl6_crii_l_53hr_647_010na_z7d5um_75msec_5POVLP_ch00_centers_intensity_radius.p'
load_range = (400,450)
dilation_radius = 5
src = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20170115_tp_bl6_lob6a_1000r_02/full_sizedatafld/20170115_tp_bl6_lob6a_1000r_647_010na_z7d5um_125msec_10povlp_ch00'
center_intensity_radius = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20170115_tp_bl6_lob6a_1000r_02/cells/20170115_tp_bl6_lob6a_1000r_647_010na_z7d5um_125msec_10povlp_ch00_centers_intensity_radius.p'
src = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20170115_tp_bl6_lob6a_1000r_02/cells/test'
center_intensity_radius = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20170115_tp_bl6_lob6a_1000r_02/cells/test_out.p'
load_range = (0,100)
'''
#load centers
center_intensity_radius = load_dictionary(center_intensity_radius) if type(
center_intensity_radius) == str else center_intensity_radius
centers = sorted(center_intensity_radius.keys(), key=lambda x: x[0])
#adjust and compensate for load range
if load_range:
centers = [
tuple((c[0] - load_range[0], c[1], c[2])) for c in centers
if c[0] in range(load_range[0], load_range[1])
]
#load
vol = load_tiff_folder(src, threshold=0, load_range=load_range)
vol0 = np.zeros_like(vol).astype('bool')
for c in centers:
vol0[c[0], c[1], c[2]] = True
vol0 = dilate_with_element(vol0 * 1, ball(dilation_radius)).astype('uint8')
#
return vol, vol0
#plot
#%matplotlib inline
save = '/home/wanglab/Downloads/rf'
plot_roc(save=save, **kwargs)
#apply
cnn_src = listdirfull(cnn_output); cnn_src.sort(); cnn_src = cnn_src[0]
inn = listdirfull(training_data_src); inn.sort();
raw_src = load_np(inn[0])[0]
gt = load_np(inn[0])[1]
centers = apply_random_forest(kwargs['classifier'], raw_src, cnn_src, size = (7,25,25))
#show
from tools.conv_net.functions.dilation import dilate_with_element, ball
gt = dilate_with_element(gt, ball(5))
src = np.zeros_like(gt)
for c in centers.astype('int'):
src[c[0],c[1],c[2]] = 1
src = dilate_with_element(src, ball(5))
#Sweep: <-- usually performance is not affected that much by this
for n_estimator in (10,20,50,100):
for max_depth in (5,10,20,50,100):
print('\n\n n_estimator--{}, max_depth--{}'.format(n_estimator, max_depth))
train_random_forest(tps, fps, n_estimator = n_estimator, max_depth = max_depth, cores = cores, test_size = test_size, dst = dst)
def apply_random_forest(classifier, raw_src, cnn_src, collect_cnn = False, size = (3,12,12), pad=False, cores=4, numZSlicesPerSplit=300, overlapping_planes = 20, chunks=10, maxip=0):
''' THIS IS MEMORY INEFFICIENT - SEE random_forest.py for better functions
vols.append(cnn_vol)
names.append('CNN\noverlay')
#make figure
fig = plt.figure(figsize=(6,8))
rows = len(names)
cols = 5
tick = 0
gs1 = gridspec.GridSpec(rows,cols)
gs1.update(wspace=0.0000025, hspace=0.000005)
adjust_exposure = True
for row in range(rows):
name = names[row]
im = np.copy(vols[row])#*100
step = im.shape[0] / cols
if name == 'CNN\noverlay':im[:,:,:,0] = dilate_with_element(im[:,:,:,0], ball(3))
for col in range(cols):
ax = plt.subplot(gs1[tick])
tick+=1; print tick
#ax = plt.subplot(rows,cols,tick)
if name == 'CNN\noverlay':
from skimage import exposure
immax = np.swapaxes(np.swapaxes(np.asarray([np.max(im[col*step:(col+1)*step,:,:,i], 0) for i in range(3)]),0,1),1,2)
bkgrn = skimage.exposure.equalize_hist(immax[:,:,1], nbins=20000)
ax.imshow(bkgrn,'gray', vmin=.6, vmax=(bkgrn.max()*1.0))
#modify colormap
import matplotlib as mpl
my_cmap = plt.cm.Reds(np.arange(plt.cm.RdBu.N))
my_cmap[:1,:4] = 0.0
my_cmap = mpl.colors.ListedColormap(my_cmap)
my_cmap.set_under('w')
return out.astype('uint8')
if __name__ == '__main__':
#goal is to generate noise, ask if objects are linear or spherical and exclude based on some feature
dims = (1, 500, 500, 500)
#choice of obejct
selem = ball(9)
selem = cube(9)
selem = cylinder(9, 9)
selem = np.asarray([star(9) for xx in range(9)])
#dilate
src = dilate_with_element(
generate_random_points(dims, numpoints=20)[0], selem)
lbl = ndi.label(src)
centers = ndi.measurements.center_of_mass(src, lbl[0],
range(1, lbl[1] + 1))
plt.imshow(np.max(src, 0))
plt.imshow(np.max(lbl[0], 0))
##
lbl_vol = np.copy(lbl[0])
lbl_vol[lbl[0] > 0] = 1
sitk.Show(sitk.GetImageFromArray(lbl_vol))
for val in range(1, lbl[1] + 1):
tvol = np.zeros_like(lbl[0])
tvol[lbl[0] == val] = 1
plt.imshow(np.max(tvol, 0))
import numpy as np
from matplotlib import rcParams
rcParams['font.family'] = 'serif'
# 3D plot of the
import matplotlib.pyplot as plt
from skimage.morphology import ball, cube, star, octagon
from tools.conv_net.functions.dilation import dilate_with_element, cylinder
import math, cv2
from mpl_toolkits.mplot3d import axes3d, Axes3D
from scipy import ndimage as ndi
if __name__ == '__main__':
#goal is to generate noise, ask if objects are linear or spherical and exclude based on some feature
dims = (1, 100, 100, 100)
src = dilate_with_element(
generate_random_points(dims, numpoints=5)[0], ball(3))
for r in range(5, 13, 2):
src = np.maximum(
src,
dilate_with_element(
generate_random_points(dims, numpoints=5)[0], ball(r)))
lbl = ndi.label(src)
centers = ndi.measurements.center_of_mass(src, lbl[0],
range(1, lbl[1] + 1))
plt.imshow(np.max(src, 0))
plt.imshow(np.max(lbl[0], 0))
val, cnts = np.unique(lbl[0], return_counts=True)
labels = size_filter(lbl, lower=200, upper=50000)
#%timeit size_filter(lbl, lower=200, upper=50000)
plt.imshow(np.max(labels[0], 0))
return np.max(src_raw[rn[0], rn[1], rn[2]])
def find_radius(centers, labels):
'''Function to return dictionary of k=centers, v=list of maximum radius/plane
'''
dct = {}
slices = find_objects(labels[0])
for i,cen in enumerate(centers):
z,y,x = [aa.astype('int') for aa in cen]
sl = slices[i]
arr = np.copy(labels[0][sl])
arr[arr!=i+1] = 0
dct[cen] = [np.max(xx).astype('float16') for xx in distance_transform_edt(arr)]
return dct
if __name__ == '__main__':
from tools.objectdetection.random_forest import apply_classifier
#dst = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20161214_db_bl6_crii_l_53hr/cells/20161214_db_bl6_crii_l_53hr_647_010na_z7d5um_75msec_5POVLP_ch00'
classifier = '/home/wanglab/wang/pisano/Python/lightsheet/supp_files/h129_rf_classifier.pkl'
cnn_src = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20161214_db_bl6_crii_l_53hr/cells/20161214_db_bl6_crii_l_53hr_647_010na_z7d5um_75msec_5POVLP_ch00_cnn_output'
raw_src = '/home/wanglab/wang/pisano/tracing_output/antero_4x/20161214_db_bl6_crii_l_53hr/full_sizedatafld/20161214_db_bl6_crii_l_53hr_647_010na_z7d5um_75msec_5POVLP_ch00'
out = apply_classifier(classifier, raw_src, cnn_src, size = (7,50,50), pad=False, cores=10, numZSlicesPerSplit=25, overlapping_planes = 10, verbose=True)
#check
from tools.conv_net.functions.dilation import dilate_with_element, generate_arr_from_pnts_and_dims, ball
pnts = [tuple((xx[0] - z, xx[1], xx[2])) for xx in out.keys()] #z in the from below eg 230
arr1 = dilate_with_element(generate_arr_from_pnts_and_dims(pnts, dims = (40, 7422, 6262)), selem=ball(5))