The aim of this project was to create our own data set consisting of images of fruits and vegetables. A subset of the data set was composed of images where the fruits and vegetables were obscured by a plastic bag. We then evaluated the difficulty of this data set using a simple kernel machine algorithm. The performance drops considerably when introducing the above mentioned subset to the data set. The algorithm was to classify the different types of fruits and vegetables present in the data set. We also created the data set in different pixel dimensions, sufficiently reducing the computation time of the algorithm while not suffering a large drop in classification performance. This enables algorithms which complexity are highly dependent on input dimension size to use the data set. From our different experimental setups we were able to conclude that the machine outperforms humans on small input dimensions, given that the humans had no prior knowledge of the data set.