TypeError: <tf.Tensor ... has type <class 'tensorflow.python.framework.ops.EagerTensor'>, but expected one of: numbers.Real

Question

I am writing a function to save images to TFRecord files in order to then read then using the Data API of TensorFlow. However, when trying to create a TFRecord to save it, I receive the following error message:

TypeError: <tf.Tensor ...> has type <class 'tensorflow.python.framework.ops.EagerTensor'>, but expected one of: numbers.Real

The function used to create the TFRecord is:

def create_tfrecord(filepath, label):
    
    image = tf.io.read_file(filepath)
    image = tf.image.decode_jpeg(image, channels=1)
    image = tf.image.convert_image_dtype(image, tf.float32)
    image = tf.image.resize(image, [299, 299])
    
    tfrecord = Example(
        features = Features(
            feature = {
                'image' : Feature(float_list=FloatList(value=[image])),
                'label' : Feature(int64_list=Int64List(value=[label]))
    })).SerializeToString()
    
    return tfrecord

If you need additional information, please let me know.

Answer 1

The problem is that image is a tensor but you need a list of float values. Try something like this:

import tensorflow as tf

def create_tfrecord(filepath, label):
    
    image = tf.io.read_file(filepath)
    image = tf.image.decode_jpeg(image, channels=1)
    image = tf.image.convert_image_dtype(image, tf.float32)
    image = tf.image.resize(image, [299, 299])
    
    tfrecord = tf.train.Example(
        features = tf.train.Features(
            feature = {
                'image' : tf.train.Feature(float_list=tf.train.FloatList(value=image.numpy().ravel().tolist())),
                'label' : tf.train.Feature(int64_list=tf.train.Int64List(value=[label]))
    })).SerializeToString()
    
    return tfrecord

create_tfrecord('/content/result_image.png', 1)

Dummy data was created like this:

import numpy
from PIL import Image

imarray = numpy.random.rand(300,300,3) * 255
im = Image.fromarray(imarray.astype('uint8')).convert('RGB')
im.save('result_image.png')

If you want to reproduce this example. When loading the tf-record, you just have to reshape the image to its original size.