def generate_preds_softmax(outputs, target_size, threshold=0.5):
preds = []
for output in outputs:
cropped = crop_image_softmax(output, target_size=target_size)
pred = binarize(cropped, threshold)
preds.append(pred)
return preds
def generate_preds(outputs, target_size):
preds = []
for output in outputs:
cropped = crop_image(output, target_size=target_size)
pred = binarize(cropped, 0.5)
preds.append(pred)
return preds
def generate_preds(args, outputs, cls_preds, target_size, threshold=0.5):
preds = []
for i, output in enumerate(outputs):
resized_img = resize_image(output, target_size=target_size)
pred = binarize(resized_img, threshold)
if args.train_cls:
pred = pred * cls_preds[i]
preds.append(pred)
return preds
def generate_preds(args, outputs, target_size, threshold=0.5):
preds = []
for output in outputs:
if args.pad_mode == 'resize':
cropped = resize_image(output, target_size=target_size)
else:
cropped = crop_image(output, target_size=target_size)
pred = binarize(cropped, threshold)
preds.append(pred)
return preds
def generate_preds(outputs, target_size, pad_mode, threshold=0.5):
preds = []
for output in outputs:
#print(output.shape)
if pad_mode == 'resize':
cropped = resize_image(output, target_size=target_size)
else:
cropped = crop_image_softmax(output, target_size=target_size)
pred = binarize(cropped, threshold)
preds.append(pred)
return preds
def predict(seed: int, porosity_request: float):
noise = generate_noise(seed, z_dim)
porosity_request = porosity_format(porosity_request)
global graph
global session
try:
with session.as_default():
with graph.as_default():
image = generator.predict([noise, porosity_request])
except BaseException as e:
print(e.args)
image = postprocessing.binarize(image[0, :, :, :, 0]).astype(np.uint8)
image = (image * 255).astype(np.uint8)
return image