CODEWITHSUNDEEP
tensorflowtensortensorflow-hub
AxelOJD
AxelOJD

Reputation: 17

Tensorflow Hub Loaded Model - ValueError: Signature specifies 343 arguments, got: 342

I'm passing a Numpy array, image, to 'process_image'. It is then processed, turned into a TensorSpec object with the dimensions and dtype required by the Tensorflow Hub model.

def image_preprocessing(image):
    img = tf.convert_to_tensor(image, dtype=tf.float32)
    img = tf.expand_dims(img, 0)
    return tf.TensorSpec.from_tensor(img)


def process_image(image):
    img = image_preprocessing(image)
    model = generate_model()

    hr_img = model(img, True)
    return hr_img[0]

img: TensorSpec(shape=(1, 480, 640, 3), dtype=tf.float32, name=None)

The model is loaded from Tensorflow Hub;

import tensorflow_hub as hub

def generate_model():
    SAVED_MODEL = 'https://tfhub.dev/captain-pool/esrgan-tf2/1'

    model = hub.load(SAVED_MODEL)

    return model

I then get this error code;

ValueError: Signature specifies 343 arguments, got: 342.

I've tried adding an additional argument (True), however it shows the exact same error as when I called model(img).

Would be thankful for any ideas.

Upvotes: 1

Views: 353

Answers (2)

Ritesh Ghorse
Ritesh Ghorse

Reputation: 181

Use tf.cast after tf.convert_to_tensor Like: img = tf.cast(tf.convert_to_tensor(image), dtype=tf.float32)

Upvotes: 0

ClaudiaR
ClaudiaR

Reputation: 3414

Try following the example of usage at esrgan-tf2:

import tensorflow_hub as hub
import tensorflow as tf
model = hub.load("https://tfhub.dev/captain-pool/esrgan-tf2/1")
# To add an extra dimension for batch, use tf.expand_dims()
dummy_image = np.ones((1, 480, 640, 3))  # [batch_size, height, width, 3]
dummy_image = tf.cast(dummy_image, tf.float32)
super_resolution = model(dummy_image) # Perform Super Resolution here

Upvotes: -1

Related Questions