SOLID TRASH SEGREGATION SYSTEM USING CONVOLUTIONAL NEURAL NETWORK

Abstract

Environmental protection has long placed a high priority on garbage sorting. Due to the cities' rapid population increase and the massive amount of waste it generates, urban regions are experiencing difficulties with their waste management systems. It is important to have an advanced waste classification system to classify a variety of solid waste materials. One of the most important steps of waste management is the segregation of the waste into the different types of components. In our country, the waste segregation process is normally done manually by hand-picking. To simplify the procedure, a trash segregation system is proposed. Waste material classification system, which is created by using the 50-layer residual network (ResNet-50) Convolutional Neural Network model which is used to categorize the waste into different types such as glass, metal, paper, and plastic, cardboard, trash. The proposed system is tested on the Kaggle Garbage dataset is able to accomplish a high accuracy. Firstly, the dataset is split into train, valid and test. Secondly, the simple CNN model and CNN based Resnet 50 model are build and use to train the training data. Before training, the image data are needed to resize and process by using data augmentation methods. Thirdly, the trained prediction model is used to classify the test data. Finally, the testing data are used to evaluate accuracy of model performance. The image with its label is coming out as output that are showed. The system is implemented by python language on google colaboratory.