CODEWITHSUNDEEP
pythontensorflowneural-network
haxtar
haxtar

Reputation: 2070

Adding a label feature using tensorflow slim

I'm using slim to convert data into TF-Record format and looking at this example, where the MNIST data-set is being converted.

On lines 127 to 128, the image png_string is assigned a label, labels[j].

example = dataset_utils.image_to_tfexample(png_string, 'png'.encode(), _IMAGE_SIZE, _IMAGE_SIZE, labels[j])

I would like to add another label, but as I look to the dataset_utils file and the image_to_tfexample function, I see:

    def image_to_tfexample(image_data, image_format, height, width, class_id):
  return tf.train.Example(features=tf.train.Features(feature={
      'image/encoded': bytes_feature(image_data),
      'image/format': bytes_feature(image_format),
      'image/class/label': int64_feature(class_id),
      'image/height': int64_feature(height),
      'image/width': int64_feature(width),
  }))

And it seems like I would have to edit this function to add another label (add another line of image/class/label': int64_feature(class_id)?)

I'm not entirely sure of how to add another label on which I'd like to train my neural network (maybe I'd just have to create another image_to_tfexample() with the same image but a different label?)

Upvotes: 0

Views: 136

Answers (1)

Vijay Mariappan
Vijay Mariappan

Reputation: 17201

Add it similar to the one you have already declared:

    def image_to_tfexample(image_data, image_format, height, width, class_id, label):
       return tf.train.Example(features=tf.train.Features(feature={
          'image/encoded': bytes_feature(image_data),
           'image/format': bytes_feature(image_format),
           'image/class/label': int64_feature(class_id),
           'image/label':int64_feature(label)
           'image/height': int64_feature(height),
           'image/width': int64_feature(width),
     }))

Remove the features you are not using, it will unnecessary increase your tfrecords size.

Upvotes: 1

Related Questions