def main(sess, ramdisk_path, image_op, predict_op):
input_size = image_op.get_shape().as_list()
print(input_size)
x_size, y_size = input_size[1:3]
PAD = int((x_size - INPUT_SIZE) / 2)
print('Working {}'.format(ramdisk_path))
svs = Slide(slide_path=ramdisk_path,
preprocess_fn=preprocess_fn,
background_speed='fast',
process_mag=PROCESS_MAG,
process_size=INPUT_SIZE,
oversample_factor=OVERSAMPLE,
verbose=True)
print('calculated foregroud: ', svs.foreground.shape)
print('calculated ds_tile_map: ', svs.ds_tile_map.shape)
svs.initialize_output('prob', dim=2, mode='tile')
PREFETCH = min(len(svs.tile_list), 1024)
def wrapped_fn(idx):
coords = svs.tile_list[idx]
img = svs._read_tile(coords) # (h, w, 3)
img = np.pad(img,
pad_width=((PAD, PAD), (PAD, PAD), (0, 0)),
mode='constant',
constant_values=0.)
return img, idx
def read_region_at_index(idx):
return tf.py_func(func=wrapped_fn,
inp=[idx],
Tout=[tf.float32, tf.int64],
stateful=False)
ds = tf.data.Dataset.from_generator(generator=svs.generate_index,
output_types=tf.int64)
ds = ds.map(read_region_at_index, num_parallel_calls=12)
ds = ds.prefetch(PREFETCH)
ds = ds.batch(BATCH_SIZE)
iterator = ds.make_one_shot_iterator()
img, idx = iterator.get_next()
print('Processing {} tiles'.format(len(svs.tile_list)))
tstart = time.time()
n_processed = 0
while True:
try:
tile, idx_ = sess.run([img, idx])
output = sess.run(predict_op, {image_op: tile})
svs.place_batch(output, idx_, 'prob', mode='tile')
n_processed += BATCH_SIZE
if n_processed % PRINT_ITER == 0:
print('[{:06d}] elapsed time [{:3.3f}] ({})'.format(
n_processed,
time.time() - tstart, tile.shape))
except tf.errors.OutOfRangeError:
print('Finished')
break
except Exception as e:
print(e)
break
dt = time.time() - tstart
spt = dt / float(len(svs.tile_list))
fps = len(svs.tile_list) / dt
print('\nFinished. {:2.2f}min {:3.3f}s/tile\n'.format(dt / 60., spt))
print('\t {:3.3f} fps\n'.format(fps))
prob_img = prob_output(svs)
svs.close()
return prob_img, fps
def process_slide(slide_path, fg_path, model, sess, out_dir, process_mag,
process_size, oversample, batch_size, n_classes):
""" Process a slide
Args:
slide_path: str
absolute or relative path to svs formatted slide
fb_path: str
absolute or relative path to foreground png
model: tfmodels.SegmentationBasemodel object
model definition to use. Weights must be restored first,
i.e. call model.restore() before passing model
sess: tf.Session
out_dir: str
path to use for output
process_mag: int
Usually one of: 5, 10, 20, 40.
Other values may work but have not been tested
process_size: int
The input size required by model.
oversample: float. Usually in [1., 2.]
How much to oversample between tiles. Larger values
will increase processing time.
batch_size: int
The batch size for inference. If the batch size is too
large given the model and process_size, then OOM errors
will be raised
n_classes: int
The number of classes output by model.
i.e. shape(model.yhat) = (batch, h, w, n_classes)
"""
print('Working {}'.format(slide_path))
# print('Working {}'.format(fg_path))
# fgimg = cv2.imread(fg_path, 0)
# fgimg = cv2.morphologyEx(fgimg, cv2.MORPH_CLOSE,
# cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (7,7)))
svs = Slide(
slide_path=slide_path,
background_speed='all',
# background_image = fgimg,
preprocess_fn=preprocess_fn,
normalize_fn=lambda x: x,
process_mag=process_mag,
process_size=process_size,
oversample=oversample,
verbose=False,
)
svs.initialize_output('prob', dim=n_classes)
svs.initialize_output('rgb', dim=3)
PREFETCH = min(len(svs.place_list), 256)
def wrapped_fn(idx):
try:
coords = svs.tile_list[idx]
img = svs._read_tile(coords)
return img, idx
except:
return 0
def read_region_at_index(idx):
return tf.py_func(func=wrapped_fn,
inp=[idx],
Tout=[tf.float32, tf.int64],
stateful=False)
ds = tf.data.Dataset.from_generator(generator=svs.generate_index,
output_types=tf.int64)
ds = ds.map(read_region_at_index, num_parallel_calls=6)
ds = ds.prefetch(PREFETCH)
ds = ds.batch(batch_size)
iterator = ds.make_one_shot_iterator()
img, idx = iterator.get_next()
print('Processing {} tiles'.format(len(svs.tile_list)))
tstart = time.time()
n_processed = 0
while True:
try:
tile, idx_ = sess.run([img, idx])
output = model.inference(tile)
svs.place_batch(output, idx_, 'prob')
svs.place_batch(tile, idx_, 'rgb')
n_processed += BATCH_SIZE
if n_processed % PRINT_ITER == 0:
print('[{:06d}] elapsed time [{:3.3f}]'.format(
n_processed,
time.time() - tstart))
except tf.errors.OutOfRangeError:
print('Finished')
dt = time.time() - tstart
spt = dt / float(len(svs.tile_list))
fps = len(svs.tile_list) / dt
print('\nFinished. {:2.2f}min {:3.3f}s/tile\n'.format(
dt / 60., spt))
print('\t {:3.3f} fps\n'.format(fps))
svs.make_outputs()
prob_img = prob_output(svs)
rgb_img = rgb_output(svs)
break
except Exception as e:
print('Caught exception at tiles {}'.format(idx_))
# print(e.__doc__)
# print(e.message)
prob_img = None
rgb_img = None
break
svs.close()
return prob_img, rgb_img, fps
def main(args):
# Translate obfuscated file names to paths if necessary
slide_list = read_list(args.f)
print('Found {} slides'.format(len(slide_list)))
encoder_args = get_encoder_args(args.encoder)
model = MilkEager(encoder_args=encoder_args,
mil_type=args.mil,
batch_size=args.batch_size,
temperature=args.temperature,
deep_classifier=args.deep_classifier)
x_pl = np.zeros((1, args.batch_size, args.input_dim, args.input_dim, 3),
dtype=np.float32)
yhat = model(tf.constant(x_pl), verbose=True)
print('yhat:', yhat.shape)
print('setting model weights')
model.load_weights(args.s, by_name=True)
## Loop over found slides:
yhats = []
for i, src in enumerate(slide_list):
print('\nSlide {}'.format(i))
basename = os.path.basename(src).replace('.svs', '')
fgpth = os.path.join(args.fg, '{}_fg.png'.format(basename))
if os.path.exists(fgpth):
ramdisk_path = transfer_to_ramdisk(
src, args.ramdisk) # never use the original src
print('Using fg image at : {}'.format(fgpth))
fgimg = cv2.imread(fgpth, 0)
try:
svs = Slide(
slide_path=ramdisk_path,
# background_speed = 'accurate',
background_speed='image',
background_image=fgimg,
preprocess_fn=lambda x: (x / 255.).astype(np.float32),
normalize_fn=lambda x: x,
process_mag=args.mag,
process_size=args.input_dim,
oversample_factor=args.oversample,
verbose=False)
except Exception as e:
print(e)
print(
'Caught SVS related error. Cleaning ramdisk and continuing.'
)
print('Cleaning file: {}'.format(ramdisk_path))
os.remove(ramdisk_path)
continue
else:
print(fgpth)
continue
svs.initialize_output(name='attention', dim=1, mode='tile')
yhat, att, indices = process_slide(svs, model, args)
yhats.append(yhat)
print('\tSlide predicted: {}'.format(yhat))
if args.mil == 'average':
print('Average MIL; continuing')
elif args.mil in ['attention', 'instance']:
print('Placing values ranged {:3.3f} - {:3.3f}'.format(
att.min(), att.max()))
print('Visualizing mean {:3.5f}'.format(np.mean(att)))
print('Visualizing std {:3.5f}'.format(np.std(att)))
svs.place_batch(att, indices, 'attention', mode='tile')
attention_img = np.squeeze(svs.output_imgs['attention'])
attention_img_raw = np.squeeze(svs.output_imgs['attention'])
attention_img = attention_img * (1. / attention_img.max())
attention_img = draw_attention(attention_img, n_bins=50)
print('attention image:', attention_img.shape, attention_img.dtype,
attention_img.min(), attention_img.max())
dst = os.path.join(args.o, '{}_att.npy'.format(basename))
np.save(dst, attention_img_raw)
dst = os.path.join(args.o, '{}_img.png'.format(basename))
cv2.imwrite(dst, attention_img)
yhat_dst = os.path.join(args.o, '{}_ypred.npy'.format(basename))
np.save(yhat_dst, yhat)
try:
svs.close()
os.remove(ramdisk_path)
del svs
except:
print('{} already removed'.format(ramdisk_path))
def process_slide(slide_path, sess, out_dir, process_mag, process_size,
oversample, batch_size, n_classes, fgspeed, fgimg):
""" Process a slide
# Pytoch - compatible mode: uses sess only for the loading pipeline
Args:
slide_path: str
absolute or relative path to svs formatted slide
model: tfmodels.SegmentationBasemodel object
model definition to use. Weights must be restored first,
i.e. call model.restore() before passing model
sess: tf.Session
out_dir: str
path to use for output
process_mag: int
Usually one of: 5, 10, 20, 40.
Other values may work but have not been tested
process_size: int
The input size required by model.
oversample: float. Usually in [1., 2.]
How much to oversample between tiles. Larger values
will increase processing time.
batch_size: int
The batch size for inference. If the batch size is too
large given the model and process_size, then OOM errors
will be raised
n_classes: int
The number of classes output by model.
i.e. shape(model.yhat) = (batch, h, w, n_classes)
"""
try:
print('Working {}'.format(slide_path))
svs = Slide(
slide_path=slide_path,
preprocess_fn=lambda x: x.astype(np.float32),
normalize_fn=lambda x: x,
process_mag=process_mag,
process_size=process_size,
oversample=oversample,
background_speed=fgspeed,
background_img=fgimg,
verbose=False,
)
# svs.initialize_output('prob', dim=n_classes)
svs.initialize_output('rgb', dim=3)
svs.initialize_output('prob', dim=n_classes, mode='tile')
PREFETCH = min(len(svs.place_list), 1024)
except Exception as e:
print(e)
print("Error loading slide at {}".format(slide_path))
return None, None, None
def wrapped_fn(idx):
try:
coords = svs.tile_list[idx]
img = svs._read_tile(coords)
return img, idx
except:
return 0
def read_region_at_index(idx):
return tf.py_func(func=wrapped_fn,
inp=[idx],
Tout=[tf.float32, tf.int64],
stateful=False)
ds = tf.data.Dataset.from_generator(generator=svs.generate_index,
output_types=tf.int64)
ds = ds.map(read_region_at_index, num_parallel_calls=8)
ds = ds.prefetch(PREFETCH)
ds = ds.batch(batch_size)
iterator = ds.make_one_shot_iterator()
img, idx = iterator.get_next()
print('Processing {} tiles'.format(len(svs.tile_list)))
tstart = time.time()
n_processed = 0
while True:
try:
tile, idx_ = sess.run([img, idx])
svs.place_batch(tile, idx_, 'rgb', clobber=True)
n_processed += BATCH_SIZE
if n_processed % PRINT_ITER == 0:
print('[{:06d}] elapsed time [{:3.3f}]'.format(
n_processed,
time.time() - tstart))
except tf.errors.OutOfRangeError:
dt = time.time() - tstart
spt = dt / float(len(svs.tile_list))
fps = len(svs.tile_list) / dt
print('\nFinished. {:2.2f}min {:3.3f}s/tile\n'.format(
dt / 60., spt))
print('{:3.3f} fps\n'.format(fps))
# svs.make_outputs()
rgb_img = rgb_output(svs)
break
except Exception as e:
print('Caught exception at tiles {}'.format(idx_))
# print(e.__doc__)
print(e)
rgb_img = None
break
svs.close()
# Replace pure black with white
bwimg = np.mean(rgb_img, axis=-1)
img_b = bwimg < 10
rgb_img[img_b, :] = 255
return rgb_img, fps
def main(args):
# Translate obfuscated file names to paths if necessary
test_list = os.path.join(args.testdir, '{}.txt'.format(args.timestamp))
test_list = read_test_list(test_list)
test_unique_ids = [
os.path.basename(x).replace('.npy', '') for x in test_list
]
if args.randomize:
np.random.shuffle(test_unique_ids)
if args.max_slides:
test_unique_ids = test_unique_ids[:args.max_slides]
slide_list, slide_labels = get_slidelist_from_uids(test_unique_ids)
print('Found {} slides'.format(len(slide_list)))
snapshot = os.path.join(args.savedir, '{}.h5'.format(args.timestamp))
# trained_model = load_model(snapshot)
# if args.mcdropout:
# encoder_args['mcdropout'] = True
encoder_args = get_encoder_args(args.encoder)
model = MilkEager(encoder_args=encoder_args,
mil_type=args.mil,
batch_size=args.batch_size,
deep_classifier=args.deep_classifier,
temperature=args.temperature)
x_pl = np.zeros((1, args.batch_size, args.input_dim, args.input_dim, 3),
dtype=np.float32)
yhat = model(tf.constant(x_pl), verbose=True)
print('yhat:', yhat.shape)
print('setting model weights')
model.load_weights(snapshot, by_name=True)
## Loop over found slides:
yhats = []
ytrues = []
for i, (src, lab) in enumerate(zip(slide_list, slide_labels)):
print('\nSlide {}'.format(i))
basename = os.path.basename(src).replace('.svs', '')
fgpth = os.path.join(args.fgdir, '{}_fg.png'.format(basename))
if os.path.exists(fgpth):
ramdisk_path = transfer_to_ramdisk(
src, args.ramdisk) # never use the original src
print('Using fg image at : {}'.format(fgpth))
fgimg = cv2.imread(fgpth, 0)
svs = Slide(
slide_path=ramdisk_path,
# background_speed = 'accurate',
background_speed='image',
background_image=fgimg,
# preprocess_fn = lambda x: (reinhard(x)/255.).astype(np.float32),
preprocess_fn=lambda x: (x / 255.).astype(np.float32),
process_mag=args.mag,
process_size=args.input_dim,
oversample_factor=args.oversample,
verbose=False)
else:
## require precomputed background; Exit.
print('Required foreground image not found ({})'.format(fgpth))
continue
svs.initialize_output(name='attention', dim=1, mode='tile')
n_tiles = len(svs.tile_list)
yhat, att, indices = process_slide(svs, model, args)
print('returned attention:', np.min(att), np.max(att), att.shape)
yhat = yhat.numpy()
yhats.append(yhat)
ytrues.append(lab)
print('\tSlide label: {} predicted: {}'.format(lab, yhat))
svs.place_batch(att, indices, 'attention', mode='tile')
attention_img = np.squeeze(svs.output_imgs['attention'])
attention_img = attention_img * (1. / attention_img.max())
attention_img = draw_attention(attention_img, n_bins=25)
print('attention image:', attention_img.shape, attention_img.dtype,
attention_img.min(), attention_img.max())
dst = os.path.join(
args.odir, args.timestamp,
'{}_{}_{:3.3f}_att.npy'.format(basename, lab, yhat[0, 1]))
np.save(dst, att)
dst = os.path.join(
args.odir, args.timestamp,
'{}_{}_{:3.3f}_img.png'.format(basename, lab, yhat[0, 1]))
cv2.imwrite(dst, attention_img)
try:
svs.close()
os.remove(ramdisk_path)
except:
print('{} already removed'.format(ramdisk_path))
yhats = np.concatenate(yhats, axis=0)
ytrues = np.array(ytrues)
acc = (np.argmax(yhats, axis=-1) == ytrues).mean()
print(acc)
def main(args):
## Search for slides
# slide_list = sorted(glob.glob(os.path.join(args.slide_dir, '*.svs')))
slide_list = read_list(args.slide_list)
print('Found {} slides'.format(len(slide_list)))
if args.shuffle:
np.random.shuffle(slide_list)
encoder_args = get_encoder_args(args.encoder)
model = ClassifierEager(encoder_args=encoder_args,
deep_classifier=True,
n_classes=args.n_classes)
fake_data = tf.constant(
np.zeros((1, args.input_dim, args.input_dim, 3), dtype=np.float32))
yhat_ = model(fake_data)
model.load_weights(args.snapshot)
model.summary()
if not os.path.exists(args.save_dir):
# shutil.rmtree(args.save_dir)
os.makedirs(args.save_dir)
## Loop over found slides:
for src in slide_list:
basename = os.path.basename(src).replace('.svs', '')
dst = os.path.join(args.save_dir, '{}.npy'.format(basename))
if os.path.exists(dst):
print('{} exists. Skipping'.format(dst))
continue
ramdisk_path = transfer_to_ramdisk(
src, args.ramdisk) # never use the original src
try:
fgpth = os.path.join(args.fgdir, '{}_fg.png'.format(basename))
fgimg = cv2.imread(fgpth, 0)
fgimg = fill_fg(fgimg)
svs = Slide(slide_path=ramdisk_path,
preprocess_fn=lambda x: (x / 255.).astype(np.float32),
normalize_fn=lambda x: x,
background_speed='image',
background_image=fgimg,
process_mag=args.mag,
process_size=args.input_dim,
oversample_factor=1.1)
svs.initialize_output(name='prob', dim=args.n_classes, mode='full')
svs.initialize_output(name='rgb', dim=3, mode='full')
n_tiles = len(svs.tile_list)
prefetch = min(512, n_tiles)
# Get tensors for image an index
iterator = get_img_idx(svs, args.batch_size, prefetch)
batches = 0
for img_, idx_ in iterator:
batches += 1
yhat = model(img_, training=False)
yhat = yhat.numpy()
idx_ = idx_.numpy()
img_ = img_.numpy()
yhat = vect_to_tile(yhat, args.input_dim)
svs.place_batch(yhat, idx_, 'prob', mode='full')
svs.place_batch(img_, idx_, 'rgb', mode='full')
if batches % 50 == 0:
print('\tbatch {:04d}'.format(batches))
svs.make_outputs(reference='prob')
prob_img = svs.output_imgs['prob']
rgb_img = svs.output_imgs['rgb'] * 255
color_img = colorize(rgb_img, prob_img)
dst = os.path.join(args.save_dir, '{}.npy'.format(basename))
np.save(dst, (prob_img * 255).astype(np.uint8))
dst = os.path.join(args.save_dir, '{}.jpg'.format(basename))
cv2.imwrite(dst, rgb_img[:, :, ::-1])
dst = os.path.join(args.save_dir, '{}_c.jpg'.format(basename))
cv2.imwrite(dst, color_img[:, :, ::-1])
except Exception as e:
print(e)
finally:
try:
print('Closing SVS')
svs.close()
except:
print('No SVS to close')
os.remove(ramdisk_path)
def main(args):
## Search for slides
slide_list = sorted(glob.glob(os.path.join(args.slide_dir, '*.svs')))
print('Found {} slides'.format(len(slide_list)))
if args.shuffle:
np.random.shuffle(slide_list)
model = ClassifierEager(encoder_args=encoder_args,
n_classes=args.n_classes)
xdummy = tf.constant(
np.zeros((args.batch_size, args.input_dim, args.input_dim, 3),
dtype=np.float32))
yhat_op = model(xdummy, verbose=True)
model.load_weights(args.snapshot, by_name=True)
# model = tf.keras.models.load_model(args.snapshot, compile=False)
model.summary()
## Loop over found slides:
for src in slide_list:
basename = os.path.basename(src).replace('.svs', '')
dst = os.path.join(args.save_dir, '{}.npy'.format(basename))
if os.path.exists(dst):
print('{} exists. continuing'.format(dst))
continue
fgpth = os.path.join(args.fgdir, '{}_fg.png'.format(basename))
if os.path.exists(fgpth):
print('fg image found:', fgpth)
fgimg = cv2.imread(fgpth, 0)
speed = 'image'
else:
speed = 'fast'
fgimg = None
ramdisk_path = transfer_to_ramdisk(
src, args.ramdisk) # never use the original src
try:
svs = Slide(slide_path=ramdisk_path,
preprocess_fn=lambda x:
(reinhard(x) / 255.).astype(np.float32),
background_speed=speed,
background_image=fgimg,
process_mag=args.mag,
process_size=args.input_dim,
oversample_factor=1.75)
svs.initialize_output(name='prob', dim=args.n_classes, mode='full')
svs.initialize_output(name='rgb', dim=3, mode='full')
n_tiles = len(svs.tile_list)
prefetch = min(128, n_tiles)
print('Tiles:', n_tiles)
# Get tensors for image an index
# img, idx = get_img_idx(svs, args.batch_size, prefetch)
iterator = get_iterator(svs, args.batch_size, prefetch)
batches = 0
for img_, idx_ in iterator:
batches += 1
# img_, idx_ = next(iterator)
yhat = model(img_, training=False)
yhat = yhat.numpy()
idx_ = idx_.numpy()
yhat = vect_to_tile(yhat, args.input_dim)
svs.place_batch(yhat, idx_, 'prob', mode='full')
svs.place_batch(img_.numpy(), idx_, 'rgb', mode='full')
if batches % 50 == 0:
print('batch {:04d}'.format(batches))
print('Making outputs')
svs.make_outputs(reference='prob')
prob_img = svs.output_imgs['prob']
print('prob img', prob_img.shape)
rgb_img = svs.output_imgs['rgb'] * 255
print('rgb img', rgb_img.shape)
color_img = colorize(rgb_img, prob_img)
print('color img', color_img.shape)
dst = os.path.join(args.save_dir, '{}.npy'.format(basename))
np.save(dst, (prob_img * 255).astype(np.uint8))
dst = os.path.join(args.save_dir, '{}.jpg'.format(basename))
cv2.imwrite(dst, rgb_img[:, :, ::-1])
dst = os.path.join(args.save_dir, '{}_c.jpg'.format(basename))
cv2.imwrite(dst, color_img[:, :, ::-1])
svs.close()
svs = []
except Exception as e:
print(e)
finally:
print('Removing {}'.format(ramdisk_path))
os.remove(ramdisk_path)
def main(args, sess):
# Translate obfuscated file names to paths if necessary
test_list = os.path.join(args.testdir, '{}.txt'.format(args.timestamp))
test_list = read_test_list(test_list)
test_unique_ids = [
os.path.basename(x).replace('.npy', '') for x in test_list
]
if args.randomize:
np.random.shuffle(test_unique_ids)
slide_list, slide_labels = get_slidelist_from_uids(test_unique_ids)
print('Found {} slides'.format(len(slide_list)))
snapshot = os.path.join(args.savedir, '{}.h5'.format(args.timestamp))
# trained_model = load_model(snapshot)
# if args.mcdropout:
# encoder_args['mcdropout'] = True
encode_model = MilkEncode(input_shape=(args.input_dim, args.input_dim, 3),
encoder_args=encoder_args,
deep_classifier=args.deep_classifier)
x_pl = tf.placeholder(shape=(None, args.input_dim, args.input_dim, 3),
dtype=tf.float32)
z_op = encode_model(x_pl)
input_shape = z_op.shape[-1]
predict_model = MilkPredict(input_shape=[input_shape],
mode=args.mil,
use_gate=args.gated_attention,
deep_classifier=args.deep_classifier)
attention_model = MilkAttention(input_shape=[input_shape],
use_gate=args.gated_attention)
print('setting encoder weights')
encode_model.load_weights(snapshot, by_name=True)
print('setting predict weights')
predict_model.load_weights(snapshot, by_name=True)
print('setting attention weights')
attention_model.load_weights(snapshot, by_name=True)
z_pl = tf.placeholder(shape=(None, input_shape), dtype=tf.float32)
y_op = predict_model(z_pl)
att_op = attention_model(z_pl)
# fig = plt.figure(figsize=(2,2), dpi=180)
## Loop over found slides:
yhats = []
ytrues = []
for i, (src, lab) in enumerate(zip(slide_list, slide_labels)):
print('\nSlide {}'.format(i))
basename = os.path.basename(src).replace('.svs', '')
fgpth = os.path.join(args.fgdir, '{}_fg.png'.format(basename))
if os.path.exists(fgpth):
ramdisk_path = transfer_to_ramdisk(
src, args.ramdisk) # never use the original src
print('Using fg image at : {}'.format(fgpth))
fgimg = cv2.imread(fgpth, 0)
svs = Slide(
slide_path=ramdisk_path,
# background_speed = 'accurate',
background_speed='image',
background_image=fgimg,
preprocess_fn=lambda x: (x / 255.).astype(np.float32),
process_mag=args.mag,
process_size=args.input_dim,
oversample_factor=args.oversample,
verbose=False)
else:
## require precomputed background; Exit.
print('Required foreground image not found ({})'.format(fgpth))
continue
svs.initialize_output(name='attention', dim=1, mode='tile')
n_tiles = len(svs.tile_list)
if not args.mcdropout:
yhat, att, indices = process_slide(svs, encode_model, y_op, att_op,
z_pl, args)
else:
yhat, att, yhat_sd, att_sd, indices = process_slide_mcdropout(
svs, encode_model, y_op, att_op, z_pl, args)
yhats.append(yhat)
ytrues.append(lab)
print('\tSlide label: {} predicted: {}'.format(lab, yhat))
svs.place_batch(att, indices, 'attention', mode='tile')
attention_img = np.squeeze(svs.output_imgs['attention'])
attention_img = attention_img * (1. / attention_img.max())
attention_img = draw_attention(attention_img, n_bins=50)
print('attention image:', attention_img.shape, attention_img.dtype,
attention_img.min(), attention_img.max())
dst = os.path.join(args.odir, args.timestamp,
'{}_att.npy'.format(basename))
np.save(dst, att)
dst = os.path.join(args.odir, args.timestamp,
'{}_img.png'.format(basename))
cv2.imwrite(dst, attention_img)
# dst = os.path.join(args.odir, args.timestamp, '{}_hist.png'.format(basename))
# fig.clf()
# plt.hist(att, bins=100);
# plt.title('Attention distribution\n{} ({} tiles)'.format(basename, n_tiles))
# plt.xlabel('Attention score')
# plt.ylabel('Tile count')
# plt.savefig(dst, bbox_inches='tight')
try:
svs.close()
os.remove(ramdisk_path)
except:
print('{} already removed'.format(ramdisk_path))
def main(args, sess):
dst = os.path.join(args.odir, 'auc_{}.png'.format(args.timestamp))
if os.path.exists(dst):
print('{} exists. Exiting.'.format(dst))
return
# Translate obfuscated file names to paths if necessary
test_list = os.path.join(args.testdir, '{}.txt'.format(args.timestamp))
test_list = read_test_list(test_list)
test_unique_ids = [
os.path.basename(x).replace('.npy', '') for x in test_list
]
slide_list, slide_labels = get_slidelist_from_uids(test_unique_ids)
print('Found {} slides'.format(len(slide_list)))
snapshot = os.path.join(args.savedir, '{}.h5'.format(args.timestamp))
trained_model = load_model(snapshot)
if args.mcdropout:
encoder_args['mcdropout'] = True
encode_model = MilkEncode(input_shape=(args.input_dim, args.input_dim, 3),
encoder_args=encoder_args)
encode_shape = list(encode_model.output.shape)
predict_model = MilkPredict(input_shape=[512],
mode=args.mil,
use_gate=args.gated_attention)
# attention_model = MilkAttention(input_shape=[512], use_gate=args.gated_attention)
models = model_utils.make_inference_functions(
encode_model,
predict_model,
trained_model,
)
encode_model, predict_model = models
z_pl = tf.placeholder(shape=(None, 512), dtype=tf.float32)
y_op = predict_model(z_pl)
fig = plt.figure(figsize=(2, 2), dpi=180)
## Loop over found slides:
yhats = []
ytrues = []
for i, (src, lab) in enumerate(zip(slide_list, slide_labels)):
print('\nSlide {}'.format(i))
basename = os.path.basename(src).replace('.svs', '')
fgpth = os.path.join(args.fgdir, '{}_fg.png'.format(basename))
if os.path.exists(fgpth):
ramdisk_path = transfer_to_ramdisk(
src, args.ramdisk) # never use the original src
print('Using fg image at : {}'.format(fgpth))
fgimg = cv2.imread(fgpth, 0)
print('fgimg shape: ', fgimg.shape)
svs = Slide(
slide_path=ramdisk_path,
# background_speed = 'accurate',
background_speed='image',
background_image=fgimg,
preprocess_fn=lambda x: (x / 255.).astype(np.float32),
process_mag=args.mag,
process_size=args.input_dim,
oversample_factor=1.5,
verbose=True)
print('calculated foregroud: ', svs.foreground.shape)
print('calculated ds_tile_map: ', svs.ds_tile_map.shape)
else:
## Require precomputed background
print('No fg image found ({})'.format(fgpth))
continue
n_tiles = len(svs.tile_list)
if not args.mcdropout:
yhat, indices = process_slide(svs, sess, encode_model, y_op, z_pl,
args)
else:
yhat, yhat_sd, indices = process_slide_mcdropout(
svs, sess, encode_model, y_op, z_pl, args)
yhats.append(yhat)
ytrues.append(lab)
print('\tSlide label: {} predicted: {}'.format(lab, yhat))
os.remove(ramdisk_path)
svs.close()
del svs
yhats = np.concatenate(yhats, axis=0)
ytrue = np.array(ytrues)
for i, yt in enumerate(ytrue):
print(yt, yhats[i, :])
auc_curve(ytrue, yhats, savepath=dst)
del encode_model
del predict_model