-4

I have nn to be trained and tested on images . I know that when I want to train nn on Image I should extract features . According to this can I consider Iris dataset which is supplied on https://archive.ics.uci.edu/ml/datasets/iris as an image dataset with extracted features and train nn on it ?? in other words can any body say to me "no you didn't train your nn on images"

thanks for help

desertnaut
  • 57,590
  • 26
  • 140
  • 166
Fahd
  • 17
  • 1
  • 6
  • 1
    No it's not image-data. – sascha Jan 08 '18 at 11:06
  • 1
    The link you provided explains the data it contains: 1. sepal length in cm 2. sepal width in cm 3. petal length in cm 4. petal width in cm 5. class: -- Iris Setosa -- Iris Versicolour -- Iris Virginica – Bhoke Jan 08 '18 at 11:08
  • I thought I could consider these data as an extracted features thanks Bhoke and Sascha – Fahd Jan 08 '18 at 11:13

1 Answers1

0

When you want to train your neural network, you have to:

  1. Preprocess your data images, in order to "clean" your dataset
  2. Decide which features you want to extract (this is a very difficult task, because you have to decide which features are relevant for recognition and which are not. This passage is already done in Iris dataset)
  3. Proceed to training and testing your neural network

So, if you look at number two, you can see that Iris dataset it's not an image dataset. However it has a number of features that you can use to train your neural network.

So the question is: Which features do you want to extract from your images in order to train your neural network?

If the answer is: the features already extracted in the Iris dataset (number of sepals, sepal width, etc.), then you can proceed to train your neural network.

If the answer is: I want to extract features directly on images, then you have to study this step.

Alex
  • 599
  • 1
  • 4
  • 16
  • fairly the features suits what I want but I'm afraid that my lecturer say I can't consider it as images with extracted features – Fahd Jan 08 '18 at 12:17
  • I'm afraid of it too, because using an external dataset means to use an external work that will be useful or not for yours, depending on the goal of your neural network. For example, if the goal is to recognise a species of Iris and these species are different given their sepal numbers, then you can use that feature (extracting it manually from image or extracting it automatically from a public dataset like Iris dataset). But if your goal is to take into account other features (like flower color or width), this database will be incomplete. – Alex Jan 08 '18 at 12:22