Marine Animal and Trash Detection using Faster RCNN

Marine debris has threatened the survival of aquatic ecosystems and has claimed the lives of several marine animals, thus disrupting the balance of nature. It is important to recognize and classify the underwater marine animals and trash in images with effective detection for their conservation and cleaning processes. This paper presents a deep learningbased method for detecting and classifying marine animals and trash from images using Faster R-CNN-a region-based convolutional neural network. It has a ResNet50 backbone with FPN, which is efficient for feature extraction and hence detects small and overlapped objects correctly. The model was trained on a custom dataset of annotated images of marine animals and different kinds of trash, which resulted in high precision and recall metrics. We have built an interactive web-based application using Streamlit that provides users with a visual and intuitive tool for real-time detection. The system showcases leveraging AI to marine conservation and cleaning and hence can also be used for environmental monitoring and research.