Oceanic Garbage Detection Using Transfer Learning and CNN

Marine pollution has become one of the most urgent environmental problems, threatening marine life and disrupting the fragile balance of ocean ecosystems. This study presents an intelligent system capable of detecting and classifying underwater waste using sophisticated machine learning techniques. The proposed framework leverages transfer learning concepts, implementing MobileNetV2 that has been fine-tuned to differentiate between four main debris types: plastic waste, metal components, glass fragments, and paper materials. The system&s;s accuracy and adaptability are improved through meticulous data preprocessing and augmentation via image transformation methods. A dedicated classification module is integrated to enable multi-class detection capabilities, with the model being trained and evaluated using an 80–20 dataset split, yielding promising results.To enhance user interaction To enhance both usability and awareness, Google Text-to-Speech is integrated into the system, offering audio messages that explain the environmental impact of each detected type of waste.A user-friendly Gradio interface The system provides a simple interface that enables users to upload images and receive instant classification results. It also underlines the role of machine learning as a practical tool for tackling environmental issues, with strong potential for future use in drones and autonomous marine vehicles to support large-scale monitoring of ocean waste.