Hierarchical deep convolutional network for automated marine plastic waste pollution detection and classification

Marine plastic waste pollution has become a serious environmental challenge affecting marine biodiversity and ecosystem sustainability. Existing marine waste monitoring approaches mainly rely on manual inspection and conventional image analysis techniques, which often suffer from limited detection accuracy and low adaptability under complex marine environmental conditions. To overcome these limitations, a Convolutional Neural Network-based binary detection framework was developed for automated identification of plastic and non-plastic marine waste. The SOUVIK dataset obtained from Kaggle, containing 2150 marine waste images, was utilized for model training and evaluation using Python and TensorFlow. The proposed framework incorporated image preprocessing, convolution operations, MaxPooling, dropout regularization, and sigmoid activation for effective feature learning and classification. Experimental evaluation achieved 88% accuracy, 88% precision, 88% recall, and 88% F1-score, outperforming several existing machine learning and deep learning approaches. The findings demonstrate the effectiveness of the proposed framework for intelligent and scalable marine pollution monitoring applications.