Automated Shells and Pebbles Classification using CNN s: A Deep Learning Approach

Using a proposed convolutional neural network model for shell and pebble recognition, this work aims at automating the classification of coastal objects. Drawn from the Kaggle database, the 865 pictures consist of two groups: stones and shells. Trained on this dataset, the proposed CNN model shows great general accuracy of 87%, therefore demonstrating its ability to differentiate between these two groups. Strong convergence with little overfitting evidenced by performance evaluation using loss and accuracy plots confirmed the resilience of the model in managing the classification task. Key performance metrics including precision, recall, and F l-scorc also revealed how consistent the model was with regard to balanced accuracy between both classes. By drastically reducing manual effort and error, this work offers a valuable foundation for the automated classification of coastal objects, hence boosting the efficiency of geological and environmental studies in coastal environments.