Classification of Carrots based on Shape Analysis using Machine Learning Techniques

In the agricultural industry, fruits and vegetables are manually graded and sorted by farmers into two categories: those which are eligible to be sold and those that are not. The most crucial aspect of this sorting is the appearance of these fruits and vegetables. They are first sorted based on shape and then on the colour. This process when done manually is prone to errors. In this work, we introduce a set of methods to automate this process of sorting by classifying carrots based on shape. The dataset is entirely built using pictures from Google Images and HOG descriptor is used for feature extraction. The first method is a simple K-Nearest Neighbours approach. The second approach uses K-Fold Cross Validation along with the first approach. Lastly the third approach implemented is a Convolutional Neural Network.