def main(args):
""" Main entry.
"""
logging.info("Creating extractor ...")
extractor = get_extractor(args.extractor, args.conf_file, {
'use_gpu': args.gpu,
'device_id': args.device_ids[0],
})
logging.info("\tDone!")
if extractor is not None:
lines = [
l.strip().split(' ') for l in open(args.image_list) if l[0] != '#'
]
heatmap_level = max(args.deep_model_level, args.heatmap_level)
for l in lines:
if len(l) > 1:
wsi, msk_image = ['%s/%s' % (args.input_folder, v) for v in l]
gen_heatmap(extractor, args.feat_name, wsi, msk_image,
args.mask_image_level, args.deep_model_level,
heatmap_level, args.window_size, args.augmentation,
args.batch_size, args.group_size, args.step_size,
args.patchmask)
else:
wsi = '%s/%s' % (args.input_folder, l[0])
gen_heatmap(extractor, args.feat_name, wsi, None,
args.mask_image_level, args.deep_model_level,
heatmap_level, args.window_size, args.augmentation,
args.batch_size, args.group_size, args.step_size,
args.patchmask)
def main(args):
""" Main entry.
"""
logging.info("Creating extractor ...")
extractor = get_extractor(
args.extractor, args.conf_file,
{
'use_gpu': args.gpu,
'device_id': args.device_ids[0],
})
logging.info("\tLoading Deep Model, Done!")
if extractor is not None:
wsi_root = '/home/dywang/CLC/Samples/wsi'
hp_imgs = [l.strip().split('/')[-1] for l in os.popen('ls %s/T*.png'%args.input_folder)]
logging.info("\tLoading Classification Model...")
classifier = load_classificer(args)
logging.info("\tLoading Classification Model, Done!")
for hp_img in hp_imgs:
hp_img_path = '%s/%s'%(args.input_folder, hp_img)
hp_img_root = hp_img.split('.')[0]
if hp_img[0] == 'T':
wsi_path = '%s/01Tumor/%s.tif'%(wsi_root, hp_img_root)
else:
wsi_path = '%s/02Normal/%s.tif'%(wsi_root, hp_img_root)
gen_heatmap_wsi(extractor, args.feat_name, classifier, wsi_path, hp_img_path, \
args.model_level, args.mask_level, args.heatmap_level, args.window_size, args.augmentation, args.batch_size, args.group_size, args.step_size)
break
def main(args):
""" Main entry.
"""
logging.info("Creating extractor ...")
extractor = get_extractor(args.extractor, args.conf_file, {
'use_gpu': args.gpu,
'device_id': args.device_ids[0],
})
logging.info("\tLoading Deep Model, Done!")
if extractor is not None:
wsi_root = '/home/dywang/CLC/Samples/wsi'
hp_imgs = []
hp_imgs += [
l.strip().split('/')[-1]
for l in os.popen('ls %s/*.png' % args.input_folder)
]
hp_imgs = hp_imgs[::-1]
for hp_img in hp_imgs:
hp_img_path = '%s/%s' % (args.input_folder, hp_img)
hp_img_root = hp_img.split('.')[0]
if hp_img_root.startswith('Tumor'):
wsi_path = '%s/01Tumor/%s.tif' % (wsi_root, hp_img_root)
elif hp_img_root.startswith("Normal"):
wsi_path = '%s/02Normal/%s.tif' % (wsi_root, hp_img_root)
elif hp_img_root.startswith("Test"):
wsi_path = '%s/03Test/%s.tif' % (wsi_root, hp_img_root)
gen_heatmap_wsi(extractor, args.feat_name, wsi_path, hp_img_path, args.threshold, \
args.model_level, args.mask_level, args.heatmap_level, args.window_size, args.augmentation, args.batch_size, args.group_size, args.step_size)
def main(args):
""" Main entry.
"""
logging.info("Creating extractor ...")
# Creates the extractor using a specified dictionary of parameters
# -> defined in the file deepfeat/__init__.py and
# -> the file deepfeat/caffe_extractor.py
extractor = get_extractor(args.extractor, args.conf_file, {
'use_gpu': args.gpu,
'device_id': args.device_ids[0],
})
logging.info("\tDone!")
# Conditions for the extractor existing:
# -> a (caffe, lasagne) model has been created
# -> a transformer model has been created
# -> the network has been successfully compiled
if extractor is not None:
# get all of the whole slide image names (stored in the argument wsi_list)
# the character '#' is used to denote comments
lines = [
l.strip().split(' ') for l in open(args.wsi_list) if l[0] != '#'
]
# The input to the program is a b/w tissue map at a particular zoom level (which is an input
# argument to the code), but it is possible to generate a heatmap of (mitoses, tumors) at any
# particular zoom level. heatmap_level is the level of the whole slide image from which the heatmap
# must be generated
heatmap_level = max(args.deep_model_level, args.heatmap_level)
print("Found " + str(len(lines)) + " lines.")
for l in lines:
if len(l) > 1:
wsi, msk_image = ['%s/%s' % (args.input_folder, v) for v in l]
gen_heatmap_wsi(extractor, args.feat_name, wsi, msk_image,
args.mask_image_level, args.deep_model_level,
heatmap_level, args.window_size,
args.augmentation, args.batch_size,
args.group_size, args.step_size)
else:
wsi = '%s/%s' % (args.input_folder, l[0])
gen_heatmap_wsi(extractor, args.feat_name, wsi, None,
args.mask_image_level, args.deep_model_level,
heatmap_level, args.window_size,
args.augmentation, args.batch_size,
args.group_size, args.step_size)
def main(args):
""" Main entry.
"""
logging.info("Creating extractor ...")
extractor = get_extractor(args.extractor, args.conf_file, {
'use_gpu': args.gpu,
'device_id': args.device_ids[0],
})
logging.info("\tDone!")
heatmap_level = max(args.deep_model_level, args.heatmap_level)
if extractor is not None:
hpgen = update_heatmap_par(args, extractor)
hpgen.run()
def extract_proc(inp_queue, extractor_name, conf_file,
img_dir, out_dir, resource):
logging.info("Creating extractor ...")
extractor = get_extractor(extractor_name, conf_file, resource)
logging.info("\tDone!")
while True:
inp = inp_queue.get()
if inp is None:
break
first_idx, v_img = inp
dump_feat(extractor, args.feat_name, v_img,
os.path.join(out_dir, "feat_%d.mat" % first_idx))
def main(args):
""" Main entry.
"""
images, first_idx, last_idx = init(args)
logging.info("Creating extractor ...")
extractor = get_extractor(args.extractor, args.conf_file, {
'use_gpu': args.gpu,
'device_id': args.device_ids[0],
})
logging.info("\tDone!")
for i in xrange(first_idx, last_idx, args.batch_size):
i_end = min(i + args.batch_size, last_idx)
v_img = load_images(args.img_dir, images, i, i_end)
dump_feat(extractor, args.feat_name, v_img,
os.path.join(args.out_dir, "feat_%010d.mat" % i))
