The previous dataset was an easy dataset for classification using texture features. In fact, many of the problems that are interesting from a business point of view are relatively easy. However, sometimes we may be faced with a tougher problem and need better and more modern techniques to get good results.

We will now test a public dataset that has the same structure: several photographs of the same class. The classes are animals, cars, transportation, and natural scenes.

When compared to the three classes' problem we discussed previously, these classes are harder to tell apart. Natural scenes, buildings, and texts have very different textures. In this dataset, however, the texture is a clear marker of the class. The following is an example from the animal class:

And here is another from the cars class:

Both objects are against natural backgrounds and with large smooth areas inside the objects. We therefore expect that textures will not be very good.

When we use the same features as before, we achieve 55 percent accuracy in cross-validation using logistic regression. This is not too bad on four classes, but not spectacular either. Let's see if we can use a different method to do better. In fact, we will see that we need to combine texture features with other methods to get the best possible results. But, first things first—we look at local features.