manipulate a matrix tensor flow, problem with tf.data.dataset

Question

I want to concatenate three images with size [1024,1024,3] to make a batch with size [3,1024,1024,3]. I wrote this code with TensorFlow but it doesn't work. It returns the error "InaccessibleTensorError: The tensor 'Tensor("truediv:0", shape=(1024, 1024, 3), dtype=float32)' cannot be accessed here: it is defined in another function or code block. Use return values, explicit Python locals or TensorFlow collections to access it.".

def decode_img(filename):
    image = tf.ones((3,1024,1024,3),dtype=tf.dtypes.float32)
    cnt=0
    slices = []
    for fi in filename:
      bits = tf.io.read_file(fi)
      img = tf.image.decode_jpeg(bits, channels=3)
      img = tf.image.resize(img, (1024,1024))
      slices.append(tf.cast(img, tf.float32) / 255.0)
      cnt +=1

    image = tf.stack(slices)
    return image

#-----------------------
filenames = ['img1.png', 'img2.png', 'img3.png']
dataset = tf.data.Dataset.from_tensor_slices(filenames)
dataset = dataset.map(decode_img, num_parallel_calls=AUTO)

Also, repeat of stackoverflow.com/questions/37697747/….

DerekG
– DerekG

2020-08-30 21:23:24 +00:00
Commented Aug 30, 2020 at 21:23 — DerekG
– DerekG, Commented Aug 30, 2020 at 21:23

DerekG · Accepted Answer · 2020-08-31 05:33:06Z

1

In general, tensorflow does not support item assignment. Rather, generate all the img layers you want and then use tf.stack() or tf.concatenate.

filename = [img1.png, img2.png, img3.png]
cnt=0
slices = []
for fi in filename:
  bits = tf.io.read_file(fi)
  img = tf.image.decode_jpeg(bits, channels=3)
  img = tf.image.resize(img, (1024,1024))
  slices.append(tf.cast(img, tf.float32) / 255.0)
  cnt +=1

image = tf.stack(slices)

edited Aug 31, 2020 at 5:33

answered Aug 30, 2020 at 21:21

DerekG

3,9781 gold badge14 silver badges27 bronze badges

Sign up to request clarification or add additional context in comments.

6 Comments

Ramin Nateghi Over a year ago

By this way, it returns error "InaccessibleTensorError: The tensor 'Tensor("truediv:0", shape=(1024, 1024, 3), dtype=float32)' cannot be accessed here: it is defined in another function or code block. Use return values, explicit Python locals or TensorFlow collections to access it"

Ramin Nateghi Over a year ago

Actually, I am using the following code, it returns the error on line "image = tf.stack(slices)" def decode_img(filename): image = tf.ones((3,1024,1024,3),dtype=tf.dtypes.float32) cnt=0 slices = [] for fi in filename: bits = tf.io.read_file(fi) img = tf.image.decode_jpeg(bits, channels=3) img = tf.image.resize(img, (1024,1024)) slices.append(tf.cast(img, tf.float32) / 255.0) cnt +=1 image = tf.stack(slices) return image filenames = [img1.png, img2.png, img3.png] dataset = tf.data.Dataset.from_tensor_slices(filenames) dataset = dataset.map(decode_img)

DerekG Over a year ago

Post in question formatted as code, this is very hard to read

DerekG Over a year ago

See, in your code you define image first, then try to assign a new value to image. Assignment is not allowed in general in tf. Every time you want to combine a few variables or modify them, you need to assign the result to a new variable. This is based on the way tensorflow defines its function graphs for efficient backpropogation I believe

Ramin Nateghi Over a year ago

sorry, that was a piece of my code, I edited the question and put the all of my code

|

Collectives™ on Stack Overflow

manipulate a matrix tensor flow, problem with tf.data.dataset

1 Answer 1

6 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related