CODEWITHSUNDEEP
tensorflow
dftidft
dftidft

Reputation: 9

Do I need to subtract RGB mean value of IMAGENET when finetune RESNET and INCEPTION privided by Tensorflow Model?

I'm new to tensorflow. I know that CAFFE needs RGB mean value subtracted in advance. But I don't see the same codes in tensorflow examples.

Do I need to subtract mean value of IMAGENET when finetuning RESNET and INCEPTION provided by Tensorflow Models?

tensorflow models

Upvotes: 0

Views: 1945

Answers (1)

Harsha Pokkalla
Harsha Pokkalla

Reputation: 1802

There are multiple ways to normalize the image. Subtract mean value of training set / normalize image to [-1,1]

In this case, they normalized every image to [-1,1] using the function tf.image.per_image_standardization() which you can see in preprocessing folder section. You can follow the same preprocessing script for Fine-Tuning as well. 

 def preprocess_for_eval(image, output_height, output_width):
      """Preprocesses the given image for evaluation.
      Args:
         image: A `Tensor` representing an image of arbitrary size.
         output_height: The height of the image after preprocessing.
         output_width: The width of the image after preprocessing.
      Returns:
         A preprocessed image.
      """
      tf.summary.image('image', tf.expand_dims(image, 0))
      # Transform the image to floats.
      image = tf.to_float(image)

      # Resize and crop if needed.
      resized_image = tf.image.resize_image_with_crop_or_pad(image,
                                                         output_width,
                                                         output_height)
      tf.summary.image('resized_image', tf.expand_dims(resized_image, 0))

      # Subtract off the mean and divide by the variance of the pixels.
      return tf.image.per_image_standardization(resized_image)

I hope this helps.

Upvotes: 1

Related Questions