def predict(in_path, out_path, win_size=30):
nuclear_channel_names = ['nuclear']
trained_network_dir = "/app/DeepCell/trained_networks/Nuclear"
prefix = "2016-07-12_nuclei_all_61x61_bn_feature_net_61x61_"
image_list = os.listdir(in_path)
models_to_use = 5
list_of_weights = [
os.path.join(trained_network_dir, prefix + str(i) + ".h5")
for i in range(models_to_use)
]
for i in range(len(image_list)):
data_location = os.path.join(in_path, image_list[i])
nuclear_location = os.path.join(in_path, image_list[i])
im = skimage.io.imread(
os.path.join(in_path, image_list[i], "nuclear.png"))
nuclear_predictions = run_models_on_directory(
data_location,
nuclear_channel_names,
nuclear_location,
model_fn=nuclear_fn,
list_of_weights=list_of_weights,
image_size_x=im.shape[1],
image_size_y=im.shape[0],
win_x=win_size,
win_y=win_size,
split=False)
def segment_SLIP_pos(data_direc, mask_direc):
data_location = data_direc
phase_location = mask_direc
phase_channel_names = ['Phase'] #['channel000']
trained_network_phase_directory = "/home/nquach/DeepCell2/trained_networks/ecoli/ecoli_all/"
phase_prefix = "2016-07-20_ecoli_all_31x31_bn_feature_net_31x31_"
#"2017-02-12_ecoli_90x_31x31_ecoli_90x_feature_net_31x31_"
win_phase = 30
image_size_x, image_size_y = get_image_sizes(data_location,
phase_channel_names)
image_size_x /= 2
image_size_y /= 2
list_of_phase_weights = []
for j in xrange(5):
phase_weights = os.path.join(trained_network_phase_directory,
phase_prefix + str(j) + ".h5")
print(phase_weights)
list_of_phase_weights += [phase_weights]
phase_predictions = run_models_on_directory(
data_location,
phase_channel_names,
phase_location,
model_fn=fn,
list_of_weights=list_of_phase_weights,
image_size_x=image_size_x,
image_size_y=image_size_y,
win_x=win_phase,
win_y=win_phase,
std=False,
split=False)
#phase_masks = segment_nuclei(phase_predictions, mask_location = mask_location, threshold = 0.75, area_threshold = 100, solidity_threshold = 0.75, eccentricity_threshold = 0.95)
print list_of_nuclear_weights
"""
Run model on directory
"""
# cytoplasm_predictions = run_models_on_directory(data_location, cyto_channel_names, cyto_location, n_features = 3, model_fn = cyto_fn,
# list_of_weights = list_of_cyto_weights, image_size_x = image_size_x, image_size_y = image_size_y,
# win_x = win_cyto, win_y = win_cyto, std = False, split = False)
nuclear_predictions = run_models_on_directory(
data_location,
nuclear_channel_names,
nuclear_location,
model_fn=nuclear_fn,
list_of_weights=list_of_nuclear_weights,
image_size_x=image_size_x,
image_size_y=image_size_y,
win_x=win_nuclear,
win_y=win_nuclear,
std=False,
split=False)
"""
Refine segmentation with active contours
"""
# nuclear_masks = segment_nuclei(img = None, color_image = True, load_from_direc = nuclear_location, mask_location = mask_location, area_threshold = 100, solidity_threshold = 0, eccentricity_threshold = 1)
# cytoplasm_masks = segment_cytoplasm(img = None, load_from_direc = cyto_location, color_image = True, nuclear_masks = nuclear_masks, mask_location = mask_location, smoothing = 1, num_iters = 120)
"""
Compute validation metrics (optional)
"""
# direc_val = os.path.join(direc_name, 'Validation')
data_location = direc_name
cyto_location = os.path.join(direc_name, 'Output')
image_size_x, image_size_y = get_image_sizes(data_location,cyto_channel_names)
#image_size_x /= 4
#image_size_y /= 4
"""
Define model
"""
list_of_cyto_weights = []
for j in xrange(1):
cyto_weights = os.path.join(trained_network_cyto_directory, cyto_prefix + str(j) + ".h5")
list_of_cyto_weights += [cyto_weights]
"""
Run model on directory
"""
cytoplasm_predictions = run_models_on_directory(data_location, cyto_channel_names, cyto_location, n_features = 3, mode\
l_fn = cyto_fn,
list_of_weights = list_of_cyto_weights, image_size_x = image_size_x, image_size_y = image_size_y,
win_x = win_cyto, win_y = win_cyto, std = False, split = True, save = \
True)
print time.clock()-t0
win = 30
image_size_x, image_size_y = get_image_sizes(data_location,
nuclear_channel_names)
"""
Define model
"""
list_of_nuclear_weights = []
for j in xrange(1):
nuclear_weights = os.path.join(trained_network_nuclear_directory,
nuclear_prefix + str(0) + ".h5")
list_of_nuclear_weights += [nuclear_weights]
"""
Run model on directory
"""
nuclear_predictions = run_models_on_directory(
data_location,
nuclear_channel_names,
nuclear_location,
n_features=3,
model_fn=nuclear_fn,
list_of_weights=list_of_nuclear_weights,
image_size_x=image_size_x,
image_size_y=image_size_y,
win_x=win,
win_y=win,
std=True,
split=True)
list_of_nuclear_weights = []
for j in xrange(5):
nuclear_weights = os.path.join(trained_network_nuclear_directory,
nuclear_prefix + str(j) + ".h5")
list_of_nuclear_weights += [nuclear_weights]
"""
Run model on directory
"""
cytoplasm_predictions = run_models_on_directory(
data_location,
cyto_channel_names,
cyto_location,
n_features=3,
model_fn=cyto_fn,
list_of_weights=list_of_cyto_weights,
image_size_x=image_size_x,
image_size_y=image_size_y,
win_x=win_cyto,
win_y=win_cyto,
std=False,
split=False)
# nuclear_predictions = run_models_on_directory(data_location, nuclear_channel_names, nuclear_location, model_fn = nuclear_fn,
# list_of_weights = list_of_nuclear_weights, image_size_x = image_size_x, image_size_y = image_size_y,
# win_x = win_nuclear, win_y = win_nuclear, std = False, split = False)
"""
Refine segmentation with active contours
"""
# nuclear_masks = segment_nuclei(img = None, color_image = True, load_from_direc = nuclear_location, mask_location = mask_location, area_threshold = 100, solidity_threshold = 0, eccentricity_threshold = 1)