Hemp Waste Classification Using Convolutional Neural Networks

Canada’s cannabis industry generates a large amount of organic waste. This waste material has the potential to be reutilized for a variety of purposes, creating economic value. It has several uses in the fields of construction, healthcare, food and energy production, textile and paper manufacturing, and the automobile industry. Grown year round, hemp is a plant material that yields a great number of seeds, fibers, and medicines. Hemp is divided into many groups according to its energy content, nutritional value, physical characteristics, and economic benefits. Making a machine learning model to classify hemp is therefore quite helpful in differentiating between them. A convolutional neural networks (CNN) based model is introduced to perform the classification of hemp waste by considering the morphological features (color and texture). Using an industrial camera, three hundred and twenty-two photos were gathered for different types of hemp waste as part of the data collecting process. Cross-validation and data augmentation were utilized to enhance this dataset. The performance of the model was assessed using accuracy as well as additional machine learning testing parameters. The model's testing accuracy score was approximately 96%. The fact that the proposed approach can be turned into a smartphone app makes it appear highly promising.