Texture Features in Prediction of Bread Edibility

The quality of the bread depends on the selection of raw ingredients and baking process. Once the bread is baked and out for selling, the edibility can be inspected using few factors as mold formation, unpleasant odor, strange taste, and hard texture. Amongst these factors, mold formation and hard texture can be captured through visual appearance. In this study, an approach is proposed to predict the edibility of bread by building a classifier utilizing texture-based feature descriptors. Also, variants of Local Binary Pattern are used for feature extraction and the performance is analyzed. It is observed from the experimentation that Opposite Color Local Binary Pattern performs well along with Random Forest Classifier with an accuracy of 90.68% on a dataset of 587 samples of bread images. The problem being addressed in this work is first of its kind with a domain of machine learning and hence is expected to open new challenges to be addressed.